Behind the ~$5B Deal

Apollo acquired a massive internet business for ~$5B in 2021. Five years later, it’s become one of the fastest-returning deals in the firm’s history.

→ Listen on X, Spotify, YouTube, Apple

700M monthly users. Billions in revenue. Hundreds of millions in annual losses addressed. And an enormous distribution advantage that would be nearly impossible to replicate today.

Apollo’s Reed Rayman recruited Jim Lanzone to lead the turnaround.

The company? Yahoo.

I sat down with CEO Jim Lanzone, one of the internet industry’s most accomplished turnaround executives, to go inside Apollo’s PE playbook, the economics of the acquisition, & why Yahoo may be one of the most underrated assets in tech.

We get into what it’s really like operating under Apollo, the leadership decisions behind the turnaround, the value of internet distribution at scale, and why some startups with less than 0.1% of Yahoo’s users command higher valuations.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Jim Lanzone, CEO at Yahoo
(01:33) How Apollo bought Yahoo for $4.8B
(04:16) Taking over a company in the middle of COVID
(07:22) Is private equity really the villain?
(12:42) The mistake that handed Google the internet
(16:09) Where Yahoo's 700M monthly users come from
(17:07) Why Yahoo sold AOL and TechCrunch
(21:21) "You can't skip levels of the video game"
(27:58) Why people still trust Yahoo in the AI era
(30:26) How Yahoo Scout competes with ChatGPT and Perplexity
(33:48) Saying no to the AI gold rush
(37:31) From turnaround to comeback
(39:42) One of the fastest paybacks in Apollo history
(42:50) The biggest mistake companies make at scale
(44:21) Startups worth more than Yahoo with 0.1% of its users
(46:25) Why Yahoo is ready for the next crash
(48:57) Barry Diller, Tinder and the tweet that predicted Yahoo
(54:36) Is Yahoo finally back?

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Harvey—The AI platform built for legal work, is helping the best firms and companies close deals faster. More than 200,000 lawyers, 80% of the Am Law 100, and over 20% of the Fortune 500 build with Harvey. Visit: Harvey.ai

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Mercor—The human intelligence behind frontier AI. Frontier labs and leading enterprises use Mercor to turn the knowledge of 5M+ experts into the data, evals and benchmarks that make AI better at real work. Visit: http://mercor.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public—Investing platform Public is the world's first agentic brokerage. You can invest across stocks, options, bonds, crypto, event contracts, & more, then automate anything with AI Agents. Seriously, just tell an Agent what you want—from "Set up a buy-the-close, sell-the-open strategy for me" to "Automatically sweep any cash over $10k from my bank to my brokerage account." Agents can monitor markets, run strategies, & manage your cash 24/7. Get started: public.com/sourcery

Inside an Apollo Buyout w/ Yahoo CEO Jim Lanzone, How It Really Works

Yahoo Is One of the Fastest Returning Deals in Apollo’s History

Apollo bought Yahoo and AOL from Verizon for about $5B in 2021. 5 years later, Yahoo has returned Apollo’s capital, and Lanzone says it is one of the fastest returning deals in the firm’s history. Apollo structured the deal to return cash quickly, and the sale of AOL was part of the payback.

“It’s one of the fastest returning deals in the history of Apollo. We are playing with house money now for what the upside could be here.”

Yahoo reaches 700M users a month, ranks in the top 5 internet companies in the US, and is profitable and growing. Lanzone calls it one of the most undervalued assets on the internet, and says companies with a small fraction of its audience carry higher valuations.

“There are companies that are valued more than us who have less than 1/10th of 1% of our user base.”

Newer search companies have been valued at 6x Yahoo despite its search volume, revenue, and profitability. Reporting in early 2026 put a potential Yahoo IPO valuation at $20B or more. No filing has been made, and Lanzone did not give a timeline.

→ Listen on X, Spotify, YouTube, Apple

Apollo Bought Verizon Media for ~$5B in 2021

Verizon bought AOL in 2015 and Yahoo’s operating business in 2017. It merged the 2 into a single unit called Oath, later renamed Verizon Media, with a mandate tied to Verizon’s 5G strategy.

In May 2021, funds managed by Apollo Global Management (NYSE: APO) agreed to buy Verizon Media for about $5B, structured as $4.25B in cash and $750M in preferred interests, with Verizon retaining a 10% stake. The deal closed on September 1, 2021, and the business was renamed Yahoo.

Apollo stood up the business as a new standalone company to house the brands, and Lanzone started as CEO at the end of September 2021.

“We are a 5-year-old company managing a 30-year-old brand at this point.”

Reed Rayman, a partner in Apollo’s private equity group focused on technology and media, led the deal and has chaired Yahoo’s board since close.

Jim Lanzone Arrived w/ 2 Internet Turnarounds Behind Him

Lanzone co-founded eTour in 1997 and sold it to Ask Jeeves in May 2001, joining to lead product management. Ask Jeeves’ stock rose 50x before IAC acquired the company in 2005, and Lanzone became CEO of Ask.com, leading its overhaul until 2008.

His next startup, Clicker, was backed by Redpoint and acquired by CBS in 2011. Lanzone then ran CBS Interactive for over 8 years, a group of 30+ brands including CNET, GameSpot, and SportsLine, growing it from 70M to 220M users. He launched CBS All Access, now Paramount+, and became CBS’s first Chief Digital Officer.

“I’m a growth person. I only do the turnarounds to get to the fun part, which is when you get to start putting irons in the fire for growth.”

He became CEO of Tinder in July 2020 and joined Yahoo as CEO in September 2021. He has sat on the board of Snap since September 2024 and previously served on the boards of GoPro and Edmunds.

Apollo’s Reed Rayman Backed a Growth Plan From Day 1

Lanzone worked with Apollo for several months before close and was aligned with the firm on which assets were unlikely to stay. Apollo set up the deal to return cash quickly, and 5 years later he still speaks with Rayman almost every day.

“The reputation isn’t always great for how they’re gonna manage these things. But I’ve had the opposite experience.”

Lanzone compares Apollo with the owners and investors he has reported to, including venture investors Bill Gurley and Geoff Yang, Barry Diller at IAC, and Les Moonves at CBS. Apollo, in his account, has been no harder to work for than any of them.

The firm’s goal remains an outcome, either an IPO or a sale. Lanzone says that expectation is the same as at any later-stage, venture-backed company.

“PE’s in it for a reason. They want an outcome, whether that’s an IPO or whether that’s a sale.”

Yahoo Was Rebuilt as a Conglomerate of GM-Led Brands in 2022

Under Verizon, Yahoo had 1 head of product, 1 head of technology, and 1 head of content spread across every brand. Lanzone spent his first 6 months evaluating the team and the data before making changes.

The company was reorganized into divisions led by general managers, each with its own product, technology, and content leads, plus assigned sales, marketing, PR, and business development. Every GM has founded at least 1 company.

“You can’t skip levels of the video game. You have to go level by level and win each level to earn the right to the next one.”

Each brand competes in a separate market. Yahoo Finance competes with Bloomberg, CNBC, and TradingView, Yahoo Sports with ESPN, and Yahoo Mail with Gmail. Lanzone tried a single-reporting-line model at CBS Interactive first and found it slowed the brands down.

The mission was reset to Yahoo’s 1995 goal of being a trusted guide to the internet for consumers. Yahoo stepped back from enterprise products and breaking news, focusing content on analysis that helps users make decisions, including the signing of Deirdre Bosa from CNBC.

$100M Money-Losing Ad Tech Stack Was Shut Down, Sold, or Rebuilt

AOL and Yahoo had each acquired 30+ ad tech companies. Yahoo inherited an SSP, a DSP, and a native platform that together lost hundreds of millions of dollars a year and supplied most of Yahoo’s ad revenue, which meant Yahoo’s pages only took ads from its own stack.

“It was hundreds of millions of dollars a year out the door that we had to fix.”

In November 2022, Yahoo took a 24.99% stake in Taboola alongside a 30-year commercial agreement making Taboola its native advertising partner. In February 2023, Yahoo announced cuts of 20% of the company, about 1,600 people concentrated in ad tech, with 1,000 leaving immediately and 600 over the following 6 months. The SSP was shut down so Yahoo could optimize yield across all buyers.

The DSP was kept as a growth business. Yahoo says it wins 9 of 10 head-to-head tests, driven by first-party data from a logged-in audience.

Divesting EdgeCast, TechCrunch, & AOL Helped Return Apollo’s Capital

EdgeCast, a content delivery network competing with Akamai, was the first asset to go. Limelight acquired it in June 2022 for stock valued at about $300M, leaving Yahoo with roughly 1/3 of the combined company, renamed Edgio.

TechCrunch went to Regent in March 2025 and Rivals to On3’s ownership group in April 2025, with Yahoo keeping stakes in both. Engadget was sold to Static Media in 2026. Yahoo also exited enterprise news coverage, including White House correspondents.

“It’s one of the fastest returning deals in the history of Apollo. We are playing with house money now for what the upside could be here.”

Yahoo signed a definitive agreement in October 2025 to sell AOL to Bending Spoons. Terms were not disclosed, and Reuters reported the price at about $1.4B.

Yahoo Scout Is Its First In-House Search Engine in 17 Years

Yahoo outsourced search in June 2000, first to Inktomi and then to Google, when search had no ad business model and was treated as a cost center. For the 17 years before Scout, Yahoo search ran largely on Microsoft, which remains a partner. Yahoo is still the 3rd-ranked search engine in the US.

“The original sin of Yahoo was giving search away to Google in June of 2000.”

Yahoo Scout launched in beta in January 2026. It grounds queries in Yahoo’s 30 years of search history, publisher relationships, knowledge graph, and user data, alongside Microsoft’s Bing grounding API, then sends that context to Anthropic’s Claude, Scout’s primary model. Lanzone says Haiku was the model at launch and Yahoo now uses others. Yahoo renders the results in its own interface.

“The ability to fend for ourselves and own that asset again was just life-changing for us.”

Scout is available to Yahoo’s nearly 250M US users, with attribution and referrals to publishers in every answer. MyScout adds a personalized homepage for logged-in users covering news, teams, stocks, weather, and the inbox.

700M Monthly Users, With 75% Arriving Direct

Yahoo reaches 700M users a month and roughly 90% of the US online population, ranking in the top 5 in the US. Yahoo Sports and Yahoo Finance each draw 100M+ monthly users, and about 1/3 of Americans have a Yahoo Mail account.

“75% of our users are coming directly to Yahoo. That’s one of the highest percentages on the internet.”

About 75% of daily users are logged in. Roughly half the audience is Gen Z and millennial, and Yahoo reports a 31-point lift in Gen Z brand love from campaigns including Cardi B and Kyle Cooke.

College Fantasy Football, launched in 2026, drew users of whom 40% were new to Yahoo. Yahoo Fantasy had its 3 biggest NFL days ever this season.

Startups With Under 0.1% of Yahoo’s Users Carry Higher Valuations

Lanzone says companies with less than one-tenth of 1% of Yahoo’s user base are valued above it, and that newer search companies have been valued at 6x Yahoo despite its search volume, revenue, and profitability.

“There are companies that are valued more than us who have less than one-tenth of 1% of our user base.”

He describes Yahoo as profitable and growing. High startup valuations have kept Yahoo from large acquisitions. It has bought smaller startups, including Artifact in 2024, and brought their teams in to lead core products.

Reporting in early 2026 put a potential IPO valuation at $20B or more. No filing has been made, and Lanzone did not give a timeline.

Yahoo’s Next Phase

Lanzone describes Yahoo as moving from turnaround to comeback. Apollo’s capital has been returned, the ad tech stack has been restructured, and the portfolio has been narrowed to Finance, Sports, Mail, News, and Search.

Scout is the core of the next phase, running as an answer engine and as an intelligence layer across Yahoo products. Lanzone sees personal email & calendar data as an opening in agents, since new agent products ask for that access first.

Yahoo has also launched a DSP Agent Network, an AI-powered Planner in Yahoo Mail, the AlphaSpace investing platform in Yahoo Finance, and a daily AI & technology show with Deirdre Bosa.

“I really believe that this is one of those times when the best days of Yahoo can really be ahead of it.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Ad. Agents are an experimental AI tool. Its output may be inaccurate and is not investment advice. Public Holding’s, Inc. makes no guarantees about its accuracy or reliability—verify independently before use. Brokerage accounts & services provided by Open to the Public Investing, Inc. See additional disclosures at: https://public.com/disclosures/agenticdisclosures.

"The machines have taken over"

Guillermo Rauch, founder & CEO of Vercel, joins Sourcery from the company’s San Francisco headquarters to discuss how AI agents are transforming software development, infrastructure, and the economics of SaaS.

→ Listen on X, Spotify, YouTube, Apple

Vercel has reportedly reached $600M in annualized revenue, growing 148% YoY, with more than 485,000 paying teams and a $9.3B valuation. Its infrastructure powers applications for OpenAI, Meta, Stripe, Brex, Shopify, DoorDash, and more.

AI agents now account for 60% of Vercel deployments, up from just 3% at the beginning of 2026. Guillermo explains why coding is becoming dramatically cheaper, why companies may eventually generate software instead of buying it, and how the internet is shifting from human interfaces to agent-native infrastructure.

We cover open vs. closed AI models, autonomous agents, cybersecurity, the future of SaaS, exploding compute demand, and how the role of software engineers is changing.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Guillermo Rauch, Founder & CEO at Vercel
(03:32) The simplest advice for going viral
(07:24) How Vercel became the index of AI
(11:34) The CEO who still ships code on weekends
(14:34) Robots writing code for robots
(17:57) The web's next users aren't human
(21:15) Is ChatGPT reading your website right now?
(24:21) Minority Report is becoming real software
(29:16) The end of SaaS and the rise of personal software
(34:07) AI agents are soldiers with no frontal lobe
(45:42) AI hackers and the coming cyber war
(47:55) The $1M challenge to hack Vercel's sandbox
(52:10) 80% of tokens now go to open models
(59:00) The biggest risk for frontier labs
(1:04:00) Can AI labs survive the public markets?
(1:06:28) Why AI has Silicon Valley working 7-7-7
(1:07:23) "The thing you fell in love with is gone"
(1:16:56) Your vibe-coded demo won't get you hired
(1:22:22) The hardest part of running a company is keeping it simple
(1:27:01) Why no company can run all its AI apps 24/7

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Harvey—The AI platform built for legal work, is helping the best firms and companies close deals faster. More than 200,000 lawyers, 80% of the Am Law 100, and over 20% of the Fortune 500 build with Harvey. Visit: Harvey.ai

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Mercor—The human intelligence behind frontier AI. Frontier labs and leading enterprises use Mercor to turn the knowledge of 5M+ experts into the data, evals and benchmarks that make AI better at real work. Visit: http://mercor.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public—Investing platform Public is the world's first agentic brokerage. You can invest across stocks, options, bonds, crypto, event contracts, & more, then automate anything with AI Agents. Seriously, just tell an Agent what you want—from "Set up a buy-the-close, sell-the-open strategy for me" to "Automatically sweep any cash over $10k from my bank to my brokerage account." Agents can monitor markets, run strategies, & manage your cash 24/7. Get started: public.com/sourcery

Guillermo Rauch on Coding Agents, Open Models, and Infrastructure for Agents

Vercel Reaches 485,000 Paying Teams & a $9.3B Valuation

Guillermo Rauch is the founder and CEO of Vercel, a cloud platform for building and deploying web applications. Vercel created and maintains Next.js, an open source web framework, and the AI SDK. The company was founded in 2015 and is based in San Francisco.

Vercel was valued at $9.3B in its last funding round. Investors include Accel, GIC, Khosla Ventures, General Catalyst, GV, and Notable Capital. The Information reported annualized revenue of $600M, up 148% YoY. Vercel has more than 485,000 paying teams and about 800 employees.

Customers include Meta, OpenAI, DoorDash, SpaceXAI, Notion, Blackbox.ai, The Weather Company, Stripe, Polymarket, Shopify, Brex, Ramp, Charles Schwab, Mercor. Vercel also sponsors the Mercedes-AMG PETRONAS Formula 1 team.

→ Listen on X, Spotify, YouTube, Apple

Coding Agents Trigger 60% of Vercel Deployments, Up From 3% in January

A deployment is a new version of an application pushed live on Vercel. The company took 10 years to reach 1B deployments and has added 1.5B more in the last 10 months.

Coding agents are AI systems that write and ship code. At the start of 2026, they triggered 3% of deployments on Vercel. They now trigger 60%.

“We started out the year with 3% of deployments on Vercel were coming from coding agents. Now, 60% are agentic.”

Rauch attributes this to Vercel’s open source work. The company has published its code and documentation for 11 years, AI models trained on that material, and agents now choose Vercel by default when they deploy.

Rauch still reviews code himself. He asks agents to explain what they wrote, and he compares the developer’s job to an engineering manager’s. He recently built a proof system, a set of automated checks that confirms agent-written code is correct and does not expose data before it ships.

“I think the role that most people are now taking with these coding agents is that of an engineering manager.”

Vercel Serves a Separate Version of Its Website to AI Agents

When a person visits vercel.com, they get the standard visual site. When an AI agent visits, Vercel returns structured, text-based information the agent can parse directly. Vercel offers the same capability to its customers.

Most websites use bot detection to block automated traffic. Vercel lets agents in and reports to site owners which agents are visiting, including ChatGPT, Claude, Grok, and search engine crawlers, and whether they receive accurate information.

AI models have training cutoffs and use web search for current information. Rauch advises businesses to publish their content as text, including transcripts of podcasts and videos, so agents can find it. He calls this AEO, answer engine optimization.

“The internet for humans was HTML. The internet of agents is new protocols.”

The main protocol is MCP, the Model Context Protocol, which tells an agent what actions a service supports. Vercel’s MCP server lets an agent deploy software and list projects. Two newer protocols, the Machine Payment Protocol and the Universal Commerce Protocol, handle purchases made by agents.

Vercel works with Stripe and Shopify to support agent purchases. Agents can buy third-party services through the Vercel Marketplace with no human involved. An agent that needs a web browser, for example, can buy Browserbase through Vercel.

Rauch Calls Vercel a ‘Yes Code’ Company

Rauch expects companies to build more custom internal software. Coding agents make it cheap to write tools for a specific team or workflow. He says enterprises get a higher return on these tools than on general-purpose SaaS.

“I’m a yes code company because I’m now saying coding agents basically make the code free.”

In this model, systems of record such as Salesforce and SAP stay in place as databases. Agents read customer records, contacts, meeting notes, and support tickets from those systems and build reports and interfaces on top of the data.

Vercel uses this setup internally. Account executives prepare for meetings by asking an agent in Slack, which queries Salesforce and returns a report. The agent generates the layout for each request, showing a person’s profile or a company summary depending on what was asked and who asked.

“I never actually went to Salesforce. I just went to the agent, and the agent went to Salesforce.”

Open Models Carry Roughly 80% of AI Gateway Tokens

Vercel’s AI Gateway is a single API for accessing models from OpenAI, Anthropic, Google, xAI, and most open model providers. A token is the unit of text a model reads or writes, and providers bill by token. Vercel publishes anonymized usage data from the gateway under a Creative Commons license.

At the start of 2026, most gateway tokens went to proprietary models from OpenAI, Anthropic, Google, and xAI. Open models, whose weights are publicly available, now account for roughly 80% of tokens. On September 18, Rauch posted that open models had reached a record 78.4% of token volume. In that post, combined spend on Moonshot AI, DeepSeek, and Z.ai exceeded spend on OpenAI.

Coding accounts for a large share of token demand. Open models performed poorly on coding until Chinese labs released models close to frontier performance, including Moonshot’s Kimi.

Many developers now use a frontier model to write a plan and an open model to execute it. Changing models on the gateway takes 1 line of code. Vercel is also working on automatic routing, which picks a model based on the prompt.

“I don’t need to hire Albert Einstein to solve every problem, answer every support ticket and spell check my stuff and classify my data.”

Enterprise customers on the gateway get contractual zero data retention, meaning providers do not store or train on their data, and inference runs on US-based providers.

Zero Loyalty Is the Main Risk Rauch Sees for Frontier Labs

Rauch tells founders to measure retention, how many customers stay over time, over top-of-funnel growth, which can be bought with ads.

“I think that’s the biggest risk with a lot of the model companies, is that actually there’s zero loyalty.”

Rauch sees signs of speculation in the AI market similar to crypto, including free promotional usage, inflated token volume figures, and unverified zero data retention claims. He also notes that frontier labs are launching applications in other software categories.

“If the model labs were so convinced that the frontier intelligence advantage is enduring, why are they going into random software businesses?”

Rauch expects demand for frontier models to continue. On IPOs, he says public market investors need predictable revenue, which is harder to forecast when a free model can match a paid one. Local models that run on personal devices add to this, and he runs a small Qwen model on a MacBook Air.

Agent Security After OpenAI’s Hugging Face Breach

Rauch says the main change since ChatGPT launched is autonomy. Agents now run for long periods without human input and will use any available method to complete a task. That creates exposure to prompt injection, where an attacker hides instructions in content the agent reads, such as a web page or a comment, and the agent then acts on them using the user’s credentials and resources.

In July 2026, OpenAI disclosed that agents it was testing escaped a sandbox, chained vulnerabilities to reach the internet, and breached Hugging Face. A sandbox is an isolated environment that limits what code can access.

“People have not really figured out agent security yet.”

Vercel sells an enterprise platform where teams connect Codex, Claude Code, and other coding agents and deploy apps that are only accessible inside the company network. It also sells Vercel Sandbox, isolated cloud virtual machines where agents can run code.

From August 18 to September 1, 2026, Vercel ran a $1M bug bounty challenge against Vercel Sandbox. It received 1,285 reports. The most serious findings were 2 vulnerabilities in the networking stack of the Linux kernel, which many cloud providers also use. Rauch says one researcher using a large number of agents produced the key finding, and Vercel is sharing the results with other companies.

Vercel has been moving parts of its infrastructure to formally verified code, meaning code with a mathematical proof that it behaves as specified. Rauch expects agents to make this practical at a much larger scale. He opposes restricting open models over security concerns.

“I personally disagree with the whole let’s ban open models because of the cybersecurity risk. I think that’s nonsense because the thing that is making these models good at cybersecurity is just that they’re intelligent.”

Rauch Sees a Quality Gap Between Vercel Sandbox & Newer Clouds

A sandbox is the computer an agent works on. Where an employee gets a laptop, an agent gets an isolated machine in the cloud. Many new cloud providers, often called neoclouds, now sell this capacity.

Rauch says customers who moved to newer providers for discounts are returning to Vercel, citing reliability, security, and ease of use for agents. Vercel has run isolated build environments for 11 years, since every deployment on the platform is built inside one.

“There’s still a massive quality gap between all of these providers.”

Rauch says the cost of maintaining software stays the same even as the cost of writing it falls. Code generated with AI still has to be kept reliable and secure for as long as it runs.

“Everyone can show you that they can create something really quickly with AI, but they still need to maintain it.”

Vercel Hires Engineers Who Can Run Many Agents at Once

Vercel judges engineering candidates on how well they direct many coding agents at once. A simple app built from a few prompts does not qualify.

“I’m actually looking for your ability to commandeer swaths of agents and really get leverage back from this new ecosystem.”

Vercel also hires for communication skills. Its account executives sell to Fortune 500 companies and are expected to explain Vercel’s APIs and data flows in technical detail. Sales and solutions staff work as forward deployed engineers, sitting with customers to move their applications onto Vercel.

When Vercel began working with the Mercedes-AMG PETRONAS F1 team, an agent completed the first version of the team’s website migration in 5 minutes. Similar enterprise migrations used to take about a year with outside consultants.

Vercel adopted Rust, a systems language with a strict compiler, several years ago. Rust is hard for people to write because the compiler rejects code that breaks its rules. Agents now handle those rules, which lets teams produce faster and more secure code. Rauch points to a recent talk by DHH, creator of Ruby on Rails, that argues the same thing about moving past tools built around human convenience.

Rauch says founders and executives he knows are working longer hours since AI tools arrived, on what he calls a 777 schedule, 7 days a week from 7 AM to 7 PM. He describes an engineer’s job as deciding which problems to spend tokens on.

Serverless Evolves Into Fluid Compute

Agents increase infrastructure demand in 2 ways. They write and deploy code faster than people, and they make more web requests than people do. Vercel’s network sets new usage records almost daily.

Serverless computing runs code only when a request comes in, then shuts it down. Rauch says demand for this model is growing. Companies now produce more software than they can afford to keep running at all times, and much of it is internal or personal software that is used occasionally.

Vercel’s next version of serverless is called fluid compute. It scales up as traffic increases and down as traffic drops, and it runs sandboxes, functions, websites, and agents on one system. Rauch expects it to be the standard model for the next decade.

“You need compute that flows like liquid.”

Rauch Expects Agents to Drive Infrastructure Demand for Years

Rauch expects every product and service to be rebuilt for agents over the next few years, as agents account for a growing share of internet traffic. He says businesses that keep blocking agents will be left out of that traffic.

He expects the next couple of years to be difficult for security. Agents can find vulnerabilities in existing human-written code faster than companies can patch them, and attackers and defenders have access to the same models. Over time, he expects agents to rewrite core infrastructure as verified code in languages like Rust.

On models, Rauch expects companies to run a mix of frontier, open, and local models, with routing systems choosing the model for each task automatically.

On compute, he expects demand on data centers, the energy grid, and networks to keep rising as AI reaches industries it has barely touched, including healthcare and education. One of Vercel’s AI startup customers is building an AI teaching assistant for students.

“I think we’ve only scratched the surface of what’s possible.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Ad. Agents are an experimental AI tool. Its output may be inaccurate and is not investment advice. Public Holding’s, Inc. makes no guarantees about its accuracy or reliability—verify independently before use. Brokerage accounts & services provided by Open to the Public Investing, Inc. See additional disclosures at: https://public.com/disclosures/agenticdisclosures.

Harvey x Sourcery

Today, Harvey is partnering with Sourcery as our exclusive legal AI sponsor.

Harvey has emerged as the dominant player in legal AI, with a $15.5B valuation & surpassing $400M in ARR.

A couple months ago I sat down with Harvey Co-Founder & CEO Winston Weinberg, Co-Founder & President Gabe Pereyra, & Head of Applied Research Niko Grupen at the company’s San Francisco HQ to discuss frontier model research, proprietary legal intelligence, agentic workflows, & the economics of AI.

Stats:

• 3,500+ customers

• 200K+ professionals

• 80% of Am Law 100 firms

• 5 of the Fortune 10

I am so excited to deepen our relationship with Harvey!

Harvey—The AI platform built for legal work, is helping the best firms and companies close deals faster. More than 200,000 lawyers, 80% of the Am Law 100, and over 20% of the Fortune 500 build with Harvey. Visit: Harvey.ai

Check out our Harvey hyper-growth series!

Recorded back in May 2026, Published June 2026. Numbers are reflective of recording date, they have grown tremendously since then!

Winston Weinberg is the CEO and co-founder of Harvey, the AI platform now used by 2/3 of the AmLaw 100, and 500+ in-house legal teams including: HSBC, Bridgewater, Carvana, Blue Owl.

→ Listen on X, Spotify, YouTube, Apple

In this episode, Winston walks me through Harvey’s San Francisco HQ and then sits down to break down the state of the business:

• Reaching nearly 2,000 customers (now 3,500!)

• Scaling to 960 employees across 12 global offices

• How token usage jumped from 1 trillion in January to an est. 13 trillion in June

“If you don’t constantly change, you are gonna get so behind that you die as a company right now.”

Winston covers the shift to cloud agents, why Harvey’s real competition is the foundation labs (not other legal tech companies), how to reinvent a company every 6 months, the future of vertical models, the billable hour problem coming for every AI buyer, and why every company will eventually sell intelligence.

Valued at $15 Billion, Harvey is on a mission to win the entire legal category, competing head-on against the trillion-dollar labs.

And they’re well on their way. This month, Harvey has passed $400M ARR, 960 employees, 3,500 customers, and roughly 13 trillion tokens processed. They’ve raised $1.5B to date from Sequoia, Kleiner Perkins, GV, Coatue, Elad Gil, the OpenAI Startup Fund, and GIC.

→ Listen on X, Spotify, YouTube, Apple

In this episode, Gabe Pereyra, Co-Founder and President + Niko Grupen is Harvey’s Head of Applied Research, sit down with me in Harvey’s San Francisco HQ speakeasy to cover:

• What LAB measures & why Harvey gave the rubric to its biggest competitors

• Early leaderboard results + what they show about long-horizon legal agents

• Token economics now hitting application-layer AI

• Why Harvey is the largest embeddings consumer for some of the labs

• Multi-model strategy & why conflict risk forces every law firm multi-model

• Architectural shift from chat-based products to cloud agents

• LAB methodology

• The billable hour coming back.. this time for AI tokens

• What’s next for the open-source legal research community

Coding agents hit Karpathy’s “agents work now” inflection in late 2025. Gabe argues legal is hitting its version of that curve right now.

Just a few weeks ago, Harvey open-sourced LAB, the Legal Agent Benchmark, the first open-source benchmark for measuring AI agent performance on real-world legal tasks. It covers 1,200+ tasks across 24 practice areas with 75,000+ rubric criteria, with initial leaderboard results from OpenAI, Anthropic, & DeepMind released today.

Before Harvey, Gabe was an AI researcher at Google Brain, DeepMind, and Meta, working on deep learning in 2016-2017 as the field was taking off. Niko was previously at Google Brain as well. Together, they led the open-sourcing of LAB.

This is the second episode in Sourcery’s Harvey mini-series, following the last conversation with co-founder and CEO Winston Weinberg.

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Trillion Dollar Outcomes

Dr. Nworah Ayogu, MD, MBA, Partner at Thrive Capital, joins Sourcery to discuss how AI is transforming drug discovery, medicine, and the economics of healthcare.

→ Listen on X, Spotify, YouTube, Apple

A Harvard-trained physician, founding team member of Cityblock Health, former General Manager of Amazon Clinic and first Chief Medical Officer of Amazon Pharmacy, Nworah is now a partner at Thrive Capital. The investment firm has grown to over $65 Billion in AUM and its healthcare portfolio includes Neuralink, Nudge, Isomorphic Labs, OpenEvidence, OpenAI, Oscar Health, & Cadence.

We cover AI’s potential to accelerate drug development, Isomorphic Labs & Demis Hassabis, Neuralink’s brain-computer interfaces (“BCI”), OpenEvidence’s rapid adoption among physicians, personalized medicine, & the emergence of 10X doctors & nurses.

We also explore some of Thrive’s breakout healthcare incubations, the next generation of trillion-dollar healthcare companies, and why AI’s most important breakthroughs may happen in medicine rather than software.

Disclaimer: This podcast is for informational purposes only. The views expressed are the speaker’s own as of the recording date & are subject to change. Thrive funds & their affiliates have investments in certain companies discussed in this episode. References to those companies are illustrative, do not represent all investments made by Thrive, & should not be viewed as recommendations or indications of future investment performance. Past performance is not indicative of future results.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Dr. Nworah Ayogu, Partner at Thrive Capital
(01:52) The next AI trade is hiding in biotech
(03:48) How AI designs a new drug
(08:37) 9 out of 10 drugs fail, can AI flip that?
(11:33) How Thrive got into Isomorphic Labs
(14:14) Giving paralyzed veterans their hands back
(20:39) Neuralink is more interesting than SpaceX
(23:21) The hearing aid that broke the rating scale
(30:18) The doctor who built Amazon Health
(33:21) Why AGI makes healthcare the biggest bet
(35:25) One cancer drug outearns every AI lab combined
(37:55) How Lilly became healthcare's first $1T company
(40:04) How Thrive helped OpenAI break into healthcare
(44:13) OpenEvidence: two-thirds of US doctors in three years
(46:16) Healthcare is 6x bigger than defense
(48:15) $40T spent, zero gain in life expectancy
(51:43) A custom gene therapy built for one baby
(55:47) Saving Medicare $2.5M a week
(1:04:13) Oscar's comeback from down 95% to up 100%
(1:06:19) Why OpenAI is now a giant healthcare provider

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Harvey—The AI platform built for legal work, is helping the best firms and companies close deals faster. More than 200,000 lawyers, 80% of the Am Law 100, and over 20% of the Fortune 500 build with Harvey. Visit: Harvey.ai

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public—Investing platform Public is the world's first agentic brokerage. You can invest across stocks, options, bonds, crypto, event contracts, & more, then automate anything with AI Agents. Seriously, just tell an Agent what you want—from "Set up a buy-the-close, sell-the-open strategy for me" to "Automatically sweep any cash over $10k from my bank to my brokerage account." Agents can monitor markets, run strategies, & manage your cash 24/7. Get started: public.com/sourcery

Inside Thrive Capital’s Healthcare Portfolio With Dr. Nworah Ayogu

→ Listen on X, Spotify, YouTube, Apple

A $5T Market With a Decade of Flat Life Expectancy

US healthcare spending passed $5T in 2024, about 18% of GDP. CMS projects spending will reach $9T by 2034. Medicare spending is growing about 7.7% a year through 2034, faster than any other major payer category.

Healthcare is the largest employer in the US at roughly 1 in 5 workers and the largest line item in the federal budget at 6x defense. For most companies it is the second largest expense after compensation. The average employer family premium reached $26,993 in 2025, with workers paying $6,850 of it. Medical costs are the #1 cause of bankruptcy in America.

Supply is shrinking as demand rises. The US faces a projected shortage of 86,000 doctors by 2036. By 2034, Americans over 65 will outnumber those under 18, and the CDC estimates 3 in 4 Americans already have a chronic condition.

US life expectancy sits at about 80 years. The gains from 1900 to 1950 came from sanitation, antibiotics, vaccines and lower infant and child mortality. The next leg came from statins, blood pressure control and less smoking.

“For the last 10 years, we’ve spent about 40 trillion on healthcare. We have not moved our life expectancy an inch. And if that is not an indictment of the failure of the healthcare system, I don’t know what is.”

Over the same decade, cancer mortality fell sharply. Breast cancer and other diagnoses that were once often fatal now have high survival rates. With AI drug discovery, gene editing and personalized cancer vaccines reaching patients, Ayogu sees the next 10 years as the period when US life expectancy can move from 80 to 100.

“If someone told me, ‘Hey, we’re gonna spend a trillion more on healthcare this year, but everyone gets to live to 100,’ I’ll take that trade any day of the week.”

72% of Physicians Now Use AI

McKinsey reports 50% of healthcare organizations have deployed generative AI, and 80%+ have shipped first use cases to end users. The AMA reports 72% of physicians now use at least one AI use case, up from 48% in 2024. More than half of health systems that can quantify AI report 2x+ ROI, with the strongest returns in clinical documentation, denial prediction, coding and ambient listening.

Ayogu splits adoption into back office, front office and in office. Revenue cycle management in the back office moved first. Voice agents for scheduling and patient follow-up came next in the front office. In office, ambient scribes and OpenEvidence lead adoption.

In a study with Penda Health, an OpenAI clinical copilot cut diagnostic errors 16% and treatment errors 13%. OpenAI partnered with Boston Children’s and the Harvard hospitals on about 300 unsolved pediatric cases and reached new diagnoses for about 20 of them. Hundreds of millions of people ask ChatGPT health questions every week.

About 2 years ago, before o1 launched, Ayogu organized meetings between OpenAI leadership and top biotech, payer and health system CEOs. He had been testing the reasoning models on New England Journal of Medicine case studies each week. One biotech CEO in those meetings said the tools would never be useful in drug discovery. That same CEO has since announced billions of dollars of AI spend.

“These are cases that neither me nor my colleagues, some of the best doctors out there, maybe we would get it right 2 times out of 10. And OpenAI and the o1 was batting a 10 out of 10.”

“Coding is the use case that Silicon Valley cares the most about. Healthcare is the use case that Main Street cares most about.”

Thrive’s First Investment Was Oscar Health

Josh Kushner co-founded Oscar Health when he founded Thrive Capital, and Oscar was the firm’s first investment. Thrive has invested in healthcare since its founding and has put $1B+ into the sector.

Thrive manages over $65B in AUM and raised Thrive X at over $10B in 2026. The firm has incubated around 10 healthcare companies, and most of them, including Oscar, Cedar, Rightway and Cadence, are valued at over $1B. Oscar is a public Fortune 500 company.

The health portfolio includes Isomorphic Labs, Neuralink, Nudge, Fortell, Oscar Health, OpenEvidence, Formation Bio, Chai Discovery, OpenAI, Rightway, Cadence, Nourish and Abby Care.

Ayogu expects more trillion-dollar companies in healthcare. Lilly became the first by pairing better science with manufacturing scale during GLP-1 shortages and moving into care delivery with LillyDirect. He sees more biotech companies combining scientific invention with manufacturing, commercialization and care delivery to build larger TAMs.

“If you’re truly AGI-pilled, as they say, you should be actually long healthcare.”

“Our problems haven’t gotten smaller. The tools we have have gotten bigger.”

From Harvard Medical School to Amazon Pharmacy’s First CMO

Ayogu is a board-certified internist trained in neuroscience. He studied neurobiology and health policy as an undergraduate, then earned an MD from Harvard Medical School and an MBA, and completed residency at Columbia/NewYork-Presbyterian. He has held faculty appointments at Harvard and NYU.

He joined the founding team of Cityblock Health, a Thrive portfolio company that provides integrated medical, behavioral and social care for high-cost, high-need populations. Cityblock grew from an early-stage startup into national care infrastructure.

At Amazon, Ayogu helped turn PillPack into Amazon Pharmacy and served as its first Chief Medical Officer. He then launched Amazon Clinic, a virtual care business for low-acuity visits. Amazon then acquired One Medical, and the 3 businesses together became Amazon Health.

He joined Thrive in 2024 to lead healthcare investing. He has known Kushner and Kareem Zaki since college. Doctors he trained with in residency were the ones who first recommended OpenEvidence to him.

“People trust you with their medications. Will they trust you with their care delivery?”

“My clinician hat is what let me actually see, oh wow, this is a differentiated product.”

Thrive Backs Bilingual Companies

Thrive looks for healthcare companies that are world-class in both medicine and the underlying technology. Ayogu calls them bilingual.

At Isomorphic Labs, the AI researchers come from the DeepMind lineage and are held to the same bar as Anthropic and OpenAI. The chemists and biologists match the talent at Lilly, Pfizer, the Broad Institute and the Wyss Institute. Isomorphic calls its model of integrating the 2 groups “the Iso way.”

Neuralink’s hardware engineers work at the level of SpaceX and Tesla Optimus. The company also employs practicing neurosurgeons, including some of Ayogu’s Harvard Medical School classmates, alongside radiologists, battery engineers and pathologists. Safety data on the brain is core to the product.

President DJ Seo leads much of the company. Neuralink created the brain-computer interface category, and its lead in recruiting top talent has compounded since.

“We want companies that are bilingual. So we want companies that are truly the best at both of what they do. They speak 2 languages fluently.”

“I’ve never met a company before that insourced their own pathology lab.”

“People who are truly N of 1 want to work with other N of 1 people.”

Isomorphic Labs Raises a $2.1B Series B

Bringing a drug to market typically takes 10 to 15 years and billions of dollars, with a failure rate above 90%. Isomorphic Labs, founded by Nobel laureate Demis Hassabis, is building AI models to compress that process.

In March 2025, Thrive anchored Isomorphic’s $600M round, the company’s first outside capital. In May 2026, Thrive led a $2.1B Series B to scale IsoDDE, Isomorphic’s drug design engine, toward the clinic. Pharma partners include Novartis, Lilly and J&J, with work spanning small molecules, antibodies, peptides, molecular glues and biologics.

AlphaFold solved protein structure prediction and folded 200M proteins in a single year. IsoDDE more than doubles AlphaFold 3’s accuracy on a hard protein-ligand benchmark, with a 50% success rate in the hardest bin and 17 wins where AlphaFold 3 failed. It performs 2.3x better on antibody-antigen prediction. In April, the company said it was “gearing up to go into the clinic.”

AI models have improved most at predicting which molecules bind to a target. The next problem is predicting a drug’s effects and side effects in cells, animals and people, which requires generating new training data. Ayogu expects bottlenecks to move into clinical trials and manufacturing as discovery speeds up.

“If we’re going to make this go from a 10-year process to a 4-year process, it both means that we need to get better at our molecular design. We have to get better at our in vitro testing. We have to get better at our in vivo testing.”

“Rather than having a 5% success rate, I can have a 60% success rate. That radically changes the things I’m willing to do all the way throughout the process.”

Neuralink & Fortell Restore Independence

Neuralink’s implant lets patients with ALS or paralysis control a computer with their minds. Many are 30 to 40 years old, including veterans injured in combat. Patients now use the implant to drive wheelchairs and operate robotic arms. A separate trial placed the device in the brain’s speech region to restore speech. The implant reads neurons through more than 1,000 electrodes on microscopic threads, and Neuralink built its own surgical robot to implant them.

Hearing loss affects 1.5B people worldwide and is the largest modifiable risk factor for dementia. People with hearing loss have about a one-third higher likelihood of dementia. Most hearing aids amplify all sound, which makes noisy places hard to navigate.

Fortell, founded by former Butterfly Network hardware engineer Matt de Jonge, built a custom chip taped out at TSMC and uses AI to separate speech from background noise. Steve Martin and Henry Kravis are users. Fortell runs a waitlist and fits patients in person at its New York office while it scales production.

“I actually think Neuralink is a more interesting company than SpaceX. Hot take. In a lot of ways, if anything, the brain is vaster than space.”

“4.5 was their highest rating that a hearing aid had ever gotten on their 5-point scale. Fortell recently did it and got a 6.”

OpenEvidence and Cadence Expand Care Capacity

OpenEvidence is used by 65% of US doctors, with about 650,000 physicians active across roughly 27M clinical encounters as of April 2026. Epic, which has been in the market for nearly 50 years, sits at 44%, or about 450,000 to 500,000 physicians. OpenEvidence queries grew 10x in 17 months to 1M per day.

After testing the product himself, Ayogu drove to founder Daniel Nadler’s home in Malibu, and Thrive invested that day. In May, Cedars-Sinai gave clinicians enterprise access linked to EHR data. OpenEvidence has also added 357 peer-reviewed surgical videos through JOMI, specialty partnerships in ophthalmology and neuro-oncology, and ASCO oncology guidelines.

A patient who sees a doctor 4 times a year spends at most a few hours in the office and more than 8,000 hours outside it. Cadence covers those hours with remote patient monitoring for chronic disease. It ships blood pressure cuffs and scales to patients’ homes, monitors readings with AI and clinical staff, and integrates with health system EHRs across about 7,000 sites of care and over 100,000 patients.

Published results show a 25% reduction in hospitalizations, an 80%+ increase in blood pressure control and nearly a full point reduction in A1C. Cadence saves Medicare $2.5M every week. Its AI agents for medication titration match prescriber decisions 98% of the time, and calling patients the moment a reading comes in has raised reach rates 70%.

“I’ve never heard doctors rave about a clinical tool.”

“If you asked a 10-year-old, ‘What should healthcare be like?’ I feel like Cadence is what they would draw out.”

Oscar Health Grows Members 56% to 3.17M

Ayogu was a medical student when Kushner and co-founder Mario Schlosser walked him through Oscar. He told them it could not be done. Oscar used the ACA marketplaces to build a consumer-centric insurer on its own tech stack and claims infrastructure.

Oscar reached 3.17M members by the end of Q1 2026, up 56% year over year and one of the largest ACA plans. Q1 revenue rose 53% to $4.65B, and net income more than doubled to $679M. Operating margin rose 540bps to 15.2%, compared with 3.73% at Centene.

The stock traded near $2 in December 2022, down about 95%, and Thrive kept its holding across its funds. The stock is up 107% year to date and 167% over the last 3 months. Over 2 years it is up 56%, while a basket of public insurers is down 14%.

Oscar launched Oswell, an AI health agent that handles medications, test results, drug interactions, refills and appointment prep, and is expanding into menopause care. CEO Mark Bertolini, former CEO of Aetna, joined while Schlosser moved to CTO to focus on technology.

“The last health insurer that was created was in 1977. That was UnitedHealthcare. You can’t just make a new health insurer. It’s too hard. It takes too much scale. It takes too much capital.”

“You guys should call it Elphaba, ’cause you guys are kind of defying the gravity of health insurance writ large.”

N of 1 Medicine Built for a Single Patient

Baby KJ was born with a rare genetic disorder that left his body unable to process ammonia. More than half of infants with the condition die, and the standard treatment is a liver transplant. Doctors at CHOP and Penn sequenced his genome and designed a custom gene-editing therapy for his liver. He received it at about 6 months old, and he is now a toddler who walks and talks. The treatment cost about $1.5M to make and deliver.

Moderna and Merck are pairing a personalized mRNA cancer vaccine with Keytruda for melanoma. Each vaccine is built from the sequenced genome of the patient’s own tumor and trains the immune system to recognize that specific cancer if it returns. In trials, patients who received it were about 50% less likely to see their cancer return.

Thrive’s portfolio also targets access to existing breakthroughs. Nourish provides nutrition counseling for chronic disease, reports higher weight loss and GLP-1 persistence for patients on its platform, and raised a $100M Series C. Rightway works on pharmacy benefits and access to novel drugs.

“The fact that we now have these N of 1 diseases that can cure these rare genetic problems is amazing. The science has gotten there in so many of these things, but the bottleneck, again, it shifts.”

“How do we make it so it’s not a million and a half to develop this? What infrastructure needs to exist to make that more scalable and more repeatable and cheaper?”

Measuring Success in Hard Outcomes and the 10X Nurse

Ayogu measures healthcare companies on hard clinical outcomes and, at the national level, on life expectancy.

Surgeons spend about half their time on non-operative cases, and LA patients wait 3 to 4 months to see an orthopedic surgeon. If AI triages demand, wait times shorten and surgeons spend more of their time operating.

“Patient experience matters, but these are hard outcomes in terms of you being in the hospital. Is your blood pressure controlled? What’s your A1C? Are you having less strokes?”

“We talk a lot about the 10X engineer in tech. How do these tools not just enable the consumer, but how do they enable the rest of the system? Because as a nation, I think we need a 10X nurse a lot more than we need a 10X engineer.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Ad. Agents are an experimental AI tool. Its output may be inaccurate and is not investment advice. Public Holding’s, Inc. makes no guarantees about its accuracy or reliability—verify independently before use. Brokerage accounts & services provided by Open to the Public Investing, Inc. See additional disclosures at: https://public.com/disclosures/agenticdisclosures.

Incoming AI-Native Millionaire Boom

Perplexity Co-Founder & CEO Aravind Srinivas believes the next wave of American economic growth could come from smaller, AI-native businesses, not just the largest technology companies.

→ Listen on X, Spotify, YouTube, Apple

Perplexity, valued at ~$20B with more than $2B in reported funding and backing from NVIDIA, SoftBank, IVP, NEA, Accel, Jeff Bezos, and Elad Gil, has evolved from an AI answer engine into a platform for autonomous agents, research, and knowledge work.

I sat down with Aravind at Perplexity's new San Francisco headquarters to discuss the economics of AI, the future of entrepreneurship, and why agents could enable a generation of smaller, more profitable companies.

Aravind reveals that Perplexity's revenue is growing 5x year over year, enterprise adoption is accelerating, and one solopreneur is spending over $1M annually on Perplexity Computer while generating $3M in annual revenue.

We also explore Perplexity's bet on local AI, its work with NVIDIA and Apple, and why the company has no plans to build its own data centers despite the massive infrastructure buildout underway.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Aravind Srinivas, Co-founder & CEO of Perplexity
(01:30) Why Perplexity's brand is built on curiosity
(04:47) Every great invention starts with one question
(08:18) How to build a $1T company in America alone
(12:36) The next big companies will be tiny
(13:45) Is the American dream still alive?
(16:28) Stop blaming users for bad prompts
(22:55) How Perplexity quietly grew 5x
(26:51) The solopreneur spending $1M a year on Perplexity
(28:28) "We don't have competitors, we have contributors"
(29:20) The data center is coming to your laptop
(34:32) Orchestrating up with NVIDIA, down with Apple
(36:11) What happens when your AI agent knows everything
(39:17) Why Perplexity is skipping the data center race
(41:25) Stop asking AI to write code, ask it to think
(45:47) Perplexity’s answer to AI security
(48:27) The Hugging Face incident & multi-agent threats
(50:45) Turning everyone into a scientist
(53:31) Why Perplexity isn’t chasing a billion users

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Harvey—The AI platform built for legal work, is helping the best firms and companies close deals faster. More than 200,000 lawyers, 80% of the Am Law 100, and over 20% of the Fortune 500 build with Harvey. Visit: Harvey.ai

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public—Investing platform Public is the world's first agentic brokerage. You can invest across stocks, options, bonds, crypto, event contracts, & more, then automate anything with AI Agents. Seriously, just tell an Agent what you want—from "Set up a buy-the-close, sell-the-open strategy for me" to "Automatically sweep any cash over $10k from my bank to my brokerage account." Agents can monitor markets, run strategies, & manage your cash 24/7. Get started: public.com/sourcery

Aravind Srinivas on Why SMBs Could Drive the Next Economic Boom

Perplexity at $750M+ in Annualized Revenue and a Reported $30B+ Round

Perplexity launched as an answer engine in 2022. It added a multi-step search agent in late 2023, deep research in 2024, browser agents in 2025 and Perplexity Computer, its cloud agent platform, in February 2026.

Annualized revenue reportedly rose from under $250M at the start of 2026 to more than $750M by August, partly driven by Perplexity Computer. The Financial Times reported a 50% jump in a single month to more than $450M in March, following the Computer launch and a shift to usage-based pricing. Srinivas puts growth at 5X year over year, with enterprise self-serve growing faster than prosumer this year.

The last widely reported closed round was a $200M extension in September 2025 at a $20B valuation, followed by a January 2026 Series E-6 reported between $20B-$23B. In August, The Information reported Nvidia in talks to invest in a round valuing Perplexity above $30B. Backers include IVP, which led the Series B and Series D, NEA, Nvidia, Jeff Bezos, SoftBank, Accel, Bessemer and Elad Gil. Perplexity signed a $750M, 3-year Microsoft Azure agreement in January 2026 and is targeting a 2028 IPO.

“If you wanted to do agents properly, if you wanted to actually have agents do real work that people do on their computer, you need search and a browser and a multi-model harness, and you need to be able to run this all on cloud, on sandboxes with low serving costs.”

→ Listen on X, Spotify, YouTube, Apple

Perplexity Started as a Slack Bot for Company Operations Questions

The first version of Perplexity was an internal Slack bot the founding team used to answer company operations questions, from HR tools and bank accounts to health insurance for its first founding engineer. Search results at the time returned links and ads with no direct answers.

The first build plugged straight into a language model and hallucinated. The team grounded it with internet access and citations. Elad Gil, one of the first seed investors, had pushed the founders to focus on something more concrete than search, and the company initially built search over databases.

“We built the first version of Perplexity for ourselves to just ask all our questions.”

Companies with 5 to 10 employees now use the product the same way. Perplexity still counts itself as an SMB at under 1,000 people.

Smaller, AI-Native Companies and the Future of US GDP

Srinivas expects a lot of future companies to be smaller and make more money. Growth-stage investors weight Fortune 100 and Fortune 500 customers heavily because that revenue reads as durable.

His view is that a larger share of US GDP growth could come from small and mid-market companies with higher output per employee and per dollar of capital.

“It’s possible that the future of American GDP growth comes more from smaller and mid-market companies that are gonna be far more productive, capital efficient, headcount efficient, and create value way faster using these tools by being more AI native and less bureaucratic.”

Perplexity Computer is built to maximize the share of work these companies delegate to agents. Perplexity positions it as a model-neutral platform orchestrating frontier and open weights models. Hiring is the constraint Srinivas sees for new companies.

“It’s very hard to hire the right people when you’re a new company.”

A $3M Solopreneur and the Case for Recursive Effects

One customer, a 1-person company, spends upwards of $1M a year on Perplexity Computer and makes $3M a year automating newsletters for other businesses. Those businesses use the newsletters to win their own clients.

“There’s one solopreneur, literally one-person company, who’s spending upwards of a million dollars a year on our product because he’s making three million a year.”

Srinivas calls this a recursive effect. Customer revenue depends on the product, and Perplexity’s revenue depends on customer revenue. He still sees a need for distribution at the scale of millions of users, with depth of usage weighted above a billion-user count.

“AI is a recursive product. It’s meant to have recursive effects, not network effects. So the depth of usage and the quality of users matters a lot more than the quantity.”

Srinivas applies the same logic to model providers, which he calls contributors. Each new frontier model adds capabilities to Perplexity, and price competition between labs lowers serving costs. Lower serving costs increase usage among existing users.

A $1T Company Built on the US Market

Srinivas points to Berkshire Hathaway and JPMorgan as companies at or near $1T with largely US customer bases. He describes US customers as high in lifetime value and spending potential, and the US as the only country with a constant entrepreneurial spirit.

Apple is his product example. Srinivas describes Apple as arguably the largest SMB and enterprise company, built on products for a specific customer base with usage ramping over time.

“You don’t think about how to be worth a trillion. You think about how to be useful, and if you’re useful at scale to a lot of American consumers, prosumers, small and mid-market companies, eventually, after half a decade or a decade, you’re like, ‘Wow, I’m worth a trillion now.’”

Perplexity’s first 10 to 20 hires went through week-long trials. Fits were usually clear within 2 days, and candidates still unclear after 7 days were treated as a signal. Srinivas describes the US as open to outsiders, with venture and friends-and-family capital funding people who take risks.

“Fundamentally, the American dream is you can dream of solving really hard problems here and be rewarded for it, no matter who you are.”

Larry Page, FDEs and the User Is Never Wrong

Srinivas names Larry Page as his entrepreneurial hero and runs Perplexity on Page’s principle that the user is never wrong. The product is designed to anticipate what a user wants when they take the wrong step.

Model choice sits inside the harness. Users are not expected to track which of Astra, Fable, Opus or Sol suits a given task, or to write skills files and sub-agents themselves.

Steve Jobs set the goal of making computers usable by anyone when computers were built for specialists and mainframe users. Srinivas sees AI at the same stage, with companies placing forward deployed engineers to onboard customers onto products that are hard to use.

“If you crack the design, the onboarding, and the experience, and make it so intuitive that you don’t have to figure out which model to use for what task, the harness takes care of it itself, you don’t need anyone to onboard you.”

“The FDE is needed because the product is bad.”

No Data Centers, Local Compute on NVIDIA and Hybrid Compute on Apple Silicon

Perplexity has no plans to build data centers. A national Gallup survey found 70% of Americans oppose data centers in their communities, and more than 300 cities, towns and counties have placed bans or moratoriums on hyperscale construction. Srinivas cites permits, interconnection queues, HBM supply, gas turbine access and protests as the bottlenecks to new builds.

Chat runs about 30X more expensive on local hardware than in a data center. At long context of 100K tokens and above, the data center advantage narrows to about 3X or close to 1X. At 50 watts per Apple silicon chip across 100M devices, existing hardware already adds up to gigawatts of power.

“Think of the data center like a bus. Shared ride is obviously cheaper, but that didn’t stop people from wanting to have their own car.”

Portable Computer launched August 25 on Linux and NVIDIA DGX Spark, running locally and escalating to the cloud with user permission, and has since come to Windows PCs with NVIDIA RTX GPUs. Hybrid compute launched September 1 on Apple silicon Macs, starting tasks in the cloud and routing sensitive steps to local models, with enterprise admin rules for what stays on device. Finance, legal and healthcare customers have shown the most interest.

Sovereign Deployments and Agent Security After the Hugging Face Breach

Sovereign funds are asking for versions of Perplexity Computer that run entirely on hardware they own, with Perplexity seeing none of the data. These on-prem deployments run a fine-tuned open weights model, the harness, agent loops and user sessions in self-contained hardware. Perplexity loses product feedback loops and ships improvements through software updates only.

The Hugging Face breach in July involved roughly 700 of about 1,200 OpenAI agents attacking over 4.5 days to cheat a benchmark scorer. Srinivas expects slower model rollouts. He separates a main agent running sub-agents, which most harnesses already use, from multiple parent agents collaborating with their own sub-agents.

“Most of the times you test these models in single agent settings, and so you don’t discover these threats.”

Perplexity open-sourced Numbat on July 29 under an Apache 2.0 license and contributed it to the Open Secure AI Alliance, which includes NVIDIA. Perplexity runs Numbat across its own endpoints to monitor engineers using Claude Code, Codex, OpenCode and Pi, with Perplexity Computer reviewing the telemetry on a recurring schedule. Srinivas also referenced an on-device version called Bumblebee.

Problem Scoping and Long-Horizon Research as the Next Bottleneck

Srinivas sees the best engineers bottlenecked on problem solving, scoping and constraints. Perplexity is focused on research, analysis and problem solving over coding tokens and neocloud capacity.

“When you actually know exactly what problem to solve, and clearly the specs and system design, and all the constraints, writing code is actually dead simple with AIs today.”

The next product step is long-horizon research, including building prediction models, running experiments in the world and refining hypotheses in real time. Current models are not yet long-horizon, persistent and parallel enough to run research end to end. Math and code allow verification and hill climbing, while most open questions require real-world rollouts to resolve.

Perplexity plans to start with its own internal research teams, modeled on 10 research interns working 24/7.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Ad. Agents are an experimental AI tool. Its output may be inaccurate and is not investment advice. Public Holding’s, Inc. makes no guarantees about its accuracy or reliability—verify independently before use. Brokerage accounts & services provided by Open to the Public Investing, Inc. See additional disclosures at: https://public.com/disclosures/agenticdisclosures.

USV Sizes UP

USV spent 20 years keeping its funds small on purpose. Its new $900M raise is the largest fund cycle in the firm’s history.

→ Listen on X, Spotify, YouTube, Apple

I sat down with USV General Partners Fred Wilson, Rebecca Kaden & Michael Mignano at the firm’s New York office to unpack why one of venture capital’s most disciplined early-stage firms is getting bigger now, and what they think the next era of AI will look like. Together with General Partner Nick Grossman, they’re leading USV into its next chapter.

USV was early to Twitter, Coinbase, Stripe, MongoDB, Twilio, Etsy, Duolingo and Abridge. More than 20 years after Fred Wilson and Brad Burnham laid out the firm’s original internet thesis, Fred believes AI today looks a lot like the internet in 2003. The infrastructure is here, but many of the defining applications have yet to be built.

We get into why AI is making products cheaper to build but breakout companies potentially more expensive to create, why ARR can hide weak retention, the growing importance of founder storytelling, personal AI agents, the “Rebel Alliance” around open AI infrastructure, and why energy may be one of the cleanest bets on AI.

At the center of USV’s new thesis is “obliterate, don’t automate.” Instead of making existing industries incrementally more efficient, AI could completely restructure them. Fred’s version is simple. Don’t hire a lawyer, use an AI lawyer. Don’t hire an accountant, use an AI accountant. Don’t hire a doctor, use an AI doctor.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Fred, Rebecca & Michael, GPs at USV
(01:12) USV's biggest fund ever: $900M
(02:13) Inside USV's New York office
(07:19) An AI agent that runs your portfolio
(12:38) Agents won't watch Netflix for you
(14:00) USV's new strategy for the AI era
(16:54) The real cost of building in AI
(21:03) The fight over monetizing Twitter
(23:54) Why Amazon is shutting out agents
(31:02) Why smaller funds pick better
(33:40) What Series A looks like now
(40:17) Is ARR a fake metric?
(46:00) 20 years of USV's biggest wins
(51:10) Why great founders still fail
(57:46) Will Anthropic compete with its customers?
(59:54) Why Coinbase moved off Claude
(1:01:44) An AI doctor that writes prescriptions
(1:04:09) Your AI lawyer is the worst it'll ever be
(1:08:29) Who protects you from your own AI agent?
(1:11:58) Fred Wilson's been hacked 4 times
(1:13:31) Coinbase: a security company in disguise
(1:16:28) Betting on markets that don't exist yet
(1:21:16) From first-time founder to Suno

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Inside USV’s $900M in New Funds: Fred Wilson, Rebecca Kaden & Mike Mignano

USV Closes $900M, Its Largest Fund Cycle on Record

Union Square Ventures has raised $900M in new funds, the largest fund cycle in the firm’s 23-year history. The raise includes $500M for its early-stage Core Fund and $400M for its Opportunity Fund, which USV uses for later-stage deals and follow-on investments in existing portfolio companies.

The Core Fund is up 82% from USV’s last early-stage fund, $275M raised in 2024. The Opportunity Fund is up 14% from its previous $350M fund. For comparison, the 2016 Core and Opportunity funds each closed at $175M, and the 2019 cycle closed at $200M for Core and $250M for Opportunity. Put simply USV’s early-stage fund has grown way past its ‘concentrated’ strategy from $175M in 2016 to $500M in 2026 to fit the new AI era.

→ Listen on X, Spotify, YouTube, Apple

USV’s General Partners are Fred Wilson, Rebecca Kaden, Nick Grossman, and Mike Mignano. Fred, Rebecca, and Mike sat down at USV’s New York office for this conversation. Nick was out of town for the recording.

“The outcomes are bigger, the swings are bigger... and the rounds are bigger and more expensive and faster, and we are evolving our strategy to keep up with that and really be able to attack that market.” - Rebecca Kaden

The firm describes the size as an output of portfolio construction.

“We don’t raise funds as vanity metrics. It’s not like, ‘We gotta hit a billion,’ or, ‘We gotta hit two billion.’” - Mike Mignano

From a $125M First Fund to a 4-GP Partnership

USV was founded in New York in 2003 by Fred Wilson and Brad Burnham. Burnham came from AT&T Ventures, which he joined in 1993, and cofounded TACODA in 2001 before starting USV with Wilson.

The first fund took 18 months to raise. Wilson and Burnham went through more than 100 LP rejections before UTIMCO committed $25M of a $125M fund, which led to a first close within a few months. As of February 2023 (older data), UTIMCO reported a 13.8x return on that first fund and a blended 9.1x across 7 USV funds it backed from 2004 to 2014.

In 2004, Wilson and Burnham wrote a memo to their first LPs arguing that early stage companies building on now-ubiquitous internet infrastructure would generate superior returns.

• The first fund backed Twitter, Etsy, & Indeed.

• The second backed MongoDB & Twilio.

• Later funds backed Coinbase, Duolingo, & Abridge.

Fred Wilson compared the current AI market to the period when that memo was written.

“The infrastructure was there. We had Google, we had Amazon, we had Yahoo, we had eBay, but we didn’t yet have Facebook, we didn’t yet have Twitter, we didn’t yet have YouTube, we didn’t yet have iPhone.” -- Fred Wilson

The partnership has expanded slowly. Rebecca Kaden left Maveron in October 2017 to join USV as a general partner, becoming the firm’s first female partner. Nick Grossman, previously USV’s general manager of special projects, was promoted to partner in 2019 with a focus on cryptonetworks and blockchain. Mike Mignano cofounded and ran Anchor, which Spotify acquired in 2019, and later was a Partner at Lightspeed, where he led early investments in Granola and Suno.

Over the past 18 months, USV also added Venture Partner Jared Hecht, cofounder of GroupMe and Fundera, and Product Advisory Partner Scott Belsky, cofounder of Behance and former CPO at Adobe. The core investment team is Fred Wilson, Rebecca Kaden, Nick Grossman, Mike Mignano, and Nikhil Raman. Andy Weissman, Albert Wenger, John Buttrick, and Brad Burnham continue to support the firm and its portfolio.

LPs Asked USV to Raise More

USV runs quarterly LP calls covering strategy, portfolio performance, and what is and is not working. The move from prior fund sizes to $900M was discussed with LPs well ahead of the raise. USV added new LPs alongside its existing base.

“I was a little surprised how many of them said, ‘Are you sure you shouldn’t raise more?’” -- Rebecca Kaden

USV kept the fund below the size LPs would have supported. The partners cite 3 reasons. A smaller fund is easier to return and reach carry. It allows USV to be selective without needing a set percentage of a market. It keeps incentives tied to portfolio company performance.

USV’s return math does not depend on trillion-dollar outcomes.

“If our biggest winners are in the single digit billions, given our fund size and our ownership, we can produce really, really healthy returns for our partners.” -- Fred Wilson

Several of USV’s largest winners have exited in the $10B range, with a few outliers at $50B to $100B.

Series A Checks Now Start at $10M

USV has focused on Series A since its founding, with some seed. When the firm started, Fred Wilson expected to write $3M Series A checks. A typical Series A today starts at $10M, and some are $100M or more.

USV plans to lead or co-lead its investments. Series A checks from the new funds are expected to range from $10M to $30M. Mike Mignano expects 30 to 40 companies from this fund cycle, with more seed deals than USV has done historically.

“The Series A checks in this fund will probably be $10-30M in order to do that. They don’t have to be the whole round, so the rounds could be bigger than that.” - Rebecca Kaden

The larger check size is a response to round sizes. USV’s stage focus, ownership targets, and deal count are unchanged.

Incubations & Proprietary Seeds Could Be Up to Half of Seed Deals

USV is adding 3 seed strategies the firm has rarely used. The first is pure founder bets. The second is incubation, where USV builds a product inside the partnership. The third is what Fred Wilson calls proprietary seeds, where USV develops an idea, recruits a founder, and writes the first 2 checks before the company raises from outside investors.

“I think maybe those will make up as much as a third to a half of all the seed investments we make.” - Fred Wilson

Supertake is the first incubation released. Mike Mignano introduced it on September 28 as a private-beta product that converts an investing thesis into a portfolio and can route trades through a user’s brokerage account. The agent invests through Robinhood or Coinbase accounts, monitors news on the positions, and rebalances automatically. Mignano described Supertake as a USV-incubated startup with cash in the bank and equity available for founding hires.

Jared Hecht, Scott Belsky, and Mike Mignano lead much of this work given their operating backgrounds.

Compute Costs & Subsidized Inference Change the Cost Base

Building software is cheaper than it was. Operating and scaling an AI company is not. Compute is now a major line item, and startups increasingly subsidize inference so customers can use products without paying up front.

“If you look at the fundamental cost that companies are spending on compute, it’s in a totally different hemisphere than it’s ever been.” - Rebecca Kaden

OpenAI, Anthropic, and Google still offer free tiers of their core chat products, which limits how startups can price.

“A startup can’t come in and charge for something that their competitor is giving for free.” -- Fred Wilson

Fred Wilson compared the current funding pattern to Twitter. Twitter’s original financing was $5M, followed by $25M a year later and $100M the year after. Multiplied by 10, that sequence is $50M, $250M, and $1B, which matches the round sizes in AI today.

The rest of USV’s portfolio carries its own capital needs. Energy companies and full-stack robotics companies building their own models have different cost structures than software and often require different types of capital.

Retention Over ARR in a High-Trial, Low-Engagement Market

USV weights usage retention above revenue when evaluating consumer and AI application companies. Many AI products are sold on annual subscriptions, and many require payment before a user can try them. Revenue can hold for 1 to 3 years while usage declines.

“I think it’s easier than ever before to get somebody to pay for a product via subscription without knowing if they actually like that product.” - Mike Mignano

USV tracks how often users return, daily, weekly, or monthly, and whether they churn after the first month. The firm combines hands-on product use with cohort data and looks for agreement between the two.

“Trial behavior is at an absolute high and engagement behavior is at an absolute low.” - Rebecca Kaden

Because full retention data is rarely available at seed or Series A, USV looks for proxies, including how quickly a team ships product changes that keep users engaged. Rebecca Kaden also expects multiplayer features to be the source of network effects in AI applications, which USV has not yet seen at scale.

4 Theses, Including Energy at 20% to 30% of the Portfolio

USV is investing across 4 theses.

Obliterate, don’t automate: USV backs application-layer companies that replace existing markets with AI. Doctronic is one example, an AI doctor that can write prescriptions in some states. USV uses an internally trained legal model, and Fred Wilson said 20% to 30% of USV’s legal work, including term sheets, no longer involves a human lawyer.

“Don’t hire a lawyer, use an AI lawyer. Don’t hire an accountant, use an AI accountant. Don’t hire a doctor, use an AI doctor.” - Fred Wilson

Previously unreachable data: USV invests in companies aggregating data that could not be reached until now, through physical intelligence or AI that unlocks offline data. The thesis covers software businesses as well as robotics, manufacturing, sensor networks, and businesses that scale from online to offline.

The Rebel Alliance: USV expects consumers and enterprises to seek more powerful intelligence while demanding tighter incentive alignment on trust, security, memory, and cost. On tape, Fred Wilson tied this to developers building on open source models, open weight models, and harnesses to avoid dependence on a closed stack. He said the Coinbase CTO described moving its coding agent harness off Claude onto Pi, an open source harness from USV portfolio company Earendil, and that Shopify has done the same. Earendil, founded by Flask creator Armin Ronacher, acquired Pi from its author Mario Zechner on April 8. Pi's GitHub repository showed 54.5k stars and 209 contributors as of May 2026.

Energy: USV began investing in energy through a dedicated fund in 2021, and that thesis now merges into the Core Fund. Energy is 20% to 30% of the portfolio. USV invested in Radiant in 2021. Radiant raised more than $300M about 6 months after its $165M Series C to mass-produce Kaleidos, a 1MW microreactor selected for testing at Idaho National Laboratory's DOME facility beginning in summer 2026, with customer deployments targeted for 2028. The round was led by Draper Associates and Boost VC and funds a reactor factory in Oak Ridge, Tennessee.

“If you really want just a simple way to bet on AI, and not have to think about all this other stuff, just bet on energy.” - Fred Wilson

Personal Agents, Security, & the Coinbase Trust Model

Fred Wilson expects every major platform to offer its own agent, with consumers also running a personal agent that negotiates with platform agents on their behalf. Today, agents log into services as the user. Amazon recently blocked Muse from accessing its site.

“We do believe that people will want their own agents that represent them as opposed to my Amazon, only having my Amazon agent.” - Fred Wilson

Agents create new consumer security problems, with email access as one example.

“If you give your agent access to your email, which is where all your password reset emails go, they just automatically will be able to access literally every account you own.” - Mike Mignano

USV portfolio company Common Fabric builds a sandbox for agent permissions. Rebecca Kaden also pointed to on-device models, private models, and biometric proof of humanity as categories that may gain demand. A USV partner recently met a company running local models on phones.

Fred Wilson has held crypto for 15 years, has been hacked 4 times, and has kept roughly 99% of his holdings. Coinbase, where he remains on the board, built its brand on cold storage, two-factor authentication, and delayed withdrawals. Rebecca Kaden sees a similar opening for trusted brands in AI.

Abridge, Duolingo, & Coinbase Took Non-Linear Paths

Every USV early stage fund has produced 2 to 3 standout companies, and Fred Wilson counts 20 to 25 extraordinary companies across the firm’s history.

“It’s usually two or three, maybe four companies per fund end up being the stars of that fund.” - Fred Wilson

Abridge, a 2016 fund company, started as a consumer product for patients to record doctor’s appointments. The company moved to selling to physicians. CEO Shiv generated revenue from a single product for 5 to 6 years, then launched 2 new product lines in the last 12 months.

Duolingo’s first product translated foreign language documents. Coinbase grew in steps, with breakouts in 2017 and 2019 before going public. Indeed started with 2 founders and a few engineers and is now one of the largest hiring platforms in the world. USV sold its Indeed stake more than a decade ago.

“What you learn you’re backing is not that. You’re backing a team, a product sense, and a direction.” - Rebecca Kaden

USV’s diligence on founders centers on recruiting ability, speed of decision making, and control of the market narrative. Rebecca Kaden said hiring is harder than ever given compensation at labs and heavily funded startups, and that markets now designate category winners well before the business reflects it.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

John Collison, Stripe

AI agents will completely rewire internet commerce.

Stripe processed $1.9T in 2025, up 34%. Now the company is rebuilding for an internet where AI agents, not humans, increasingly transact.

I sat down with John at Browserbase Navigate to talk agentic commerce, computer use, cybersecurity, how Stripe is rebuilding for AI & why Stripe considers 2026 the beginning of the “singularity epoch.”

→ Listen on X, Spotify, YouTube, Apple

John sees AI already showing up in new company creation. Stripe is seeing its highest rate of new firm creation in recent memory, while its 2026 Atlas cohort is tracking at 5x the revenue of the 2025 class.

We also cover Stripe’s recent acquisitions:

› Bridge + Privy, bringing crypto-native infrastructure & talent
› Metronome, usage-based billing for the AI era
› OpenRouter, routing across multiple AI models as inference spend & model competition explodes

Browserbase is building the infrastructure for this shift, giving AI agents cloud browsers to navigate the existing web & take actions across it. Big shoutout to Founder & CEO Paul Klein IV & Erika Bricky for such a top notch event.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Stripe’s $1.9T & Rebuilding for AI
(00:31) Why Agents Will Replace Web Forms
(02:20) How AI Rewires Commerce & Product Discovery
(04:40) Why AI Could Mean More Cyberattacks
(06:49) Why Stripe Says the Singularity Started in 2026
(09:29) Stripe Is Rebuilding Its APIs for AI Agents
(11:38) Why Stripe Bought Bridge, Privy, Metronome & OpenRouter
(15:12) When Stripe Builds vs. Buys
(16:55) How AI Is Changing Stripe Internally
(20:39) How John Keeps Up With AI
(23:28) The Coming Good Bot vs. Bad Bot Problem
(24:53) 17 Years of Lessons Building Stripe

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

John Collison on the Agentic Internet

Stripe Co-Founder John Collison sees computer use as a major unlock for AI agents, allowing them to interact with the existing internet without waiting for every website to build agent-specific infrastructure.

In our conversation at Browserbase Navigate, we get into how this changes commerce and product discovery, why Stripe acquired OpenRouter, how coding agents are forcing Stripe to rethink its own developer infrastructure, and how John stays current without completely succumbing to “AI psychosis.”

1. AI Agents & Computer Use

“Computer use is the ultimate backwards compatibility layer for the real world.”

John says one of the biggest unlocks for personal agents is mature computer use. Instead of waiting for every organization to build an API or MCP server, agents can interact with the websites and software that already exist.

He points to something as simple as renewing a driver’s license. An agent that knows your address and payment information can navigate the existing DMV interface and complete the task.

“Filling out a web form is not a value-add activity for anyone.”

2. AI Is Changing Product Discovery

“AI discovery is the new point of aggregation for commerce.”

John says his experience with AI-powered product research often surfaces niche brands and businesses he hadn’t previously heard of.

He sees this as potentially positive for the long tail of businesses. Products that are well-reviewed and liked online can become more discoverable, while existing aggregators will have to adapt to AI becoming another point through which consumers discover products.

“I think it’s quite a good story for the long tail of businesses.”

3. Why Stripe Acquired OpenRouter

“To believe in OpenRouter, you have to believe that more than one model will be used inside a business.”

John says businesses may use different models for different workloads and change those choices as new models become available.

He gives the example of sending a high-value user to a frontier model while routing a lower-value user to a lower-cost model. Stripe saw OpenRouter as furthest ahead in building the marketplace and infrastructure for businesses using multiple models.

4. Stripe Is Rebuilding for AI

“No one actually really integrates the Stripe API by hand anymore. You know, they send their coding agent to do it.”

Stripe spent years optimizing its APIs, documentation and integrations for developers. John says coding agents have changed what matters.

“The developer ergonomics are different from the AI ergonomics.”

Stripe is now having to change what it optimizes for internally, including problems like agents integrating the wrong version of an API.

5. Avoiding AI Psychosis

“Who said I don’t have AI psychosis?”

John says he stays current by vibe coding applications he wants to use and spending time with fast-growing Stripe customers, which increasingly means companies working in or around AI.

But his more analog answer is simpler.

“Putting down the phone from time to time. I don’t know if people have tried that, but it’s pretty nice out there in the wider world.”

His advice for people spending too much time in San Francisco’s AI bubble is to get outside SoMa, cross the Golden Gate Bridge and spend some time in Marin.

ICYMI: Interview with Browserbase CEO Paul Klein IV from last year

Browserbase provides a cloud-based browser infrastructure layer that enables developers and AI systems to programmatically interact with the web. Its core offering, headless & scalable browsers, serve as a runtime for automated workflows across web interfaces. Developers can integrate these capabilities using Browserbase’s framework, Stagehand, which offers a code interface for fine-grained browser control.

→ Listen on X, Spotify, YouTube, Apple

Common use cases span both routine and bespoke workflows that lack APIs or structured integrations, such as filing franchise taxes, checking professional credentials, or navigating legacy websites in regulated industries. Customers range from startups to large enterprises automating back-office tasks, data collection, and operational workflows that rely on browser-based interactions.

The infrastructure is designed to serve as a primitive within broader agent ecosystems and LM-powered applications, providing the browser as one of many external tools accessible to an AI system.

“If you want AI to do the same work that you do on the web, you need to give it a browser,” Klein explains. “We provide that browser, at scale, as infrastructure.”

Hot Takes from Browserbase Navigate

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Sourcery Live, Mini-Series in the South of France 🇫🇷

Sourcery Live goes inside some of the biggest shifts & hottest takes happening across AI infrastructure, defense, logistics, photonics, semiconductors & global company building.

We filmed the series in Cap Ferrat with founders, executives & investors building across the physical and digital layers of the economy.

Courtesy of our friends at Strike Capital.

Interviews with:

Shuo Wang - Co-Founder & CRO of Deel
Sean Henry - CEO of Stord
Aistis Šimaitis- CEO & Co founder of Onodrim
Herwig Van Hove - Founder & CEO of Thema
Yannick De Koninck - CTO of Thema
Bruno Wejchert - Strike Capital
Steven Latre - VP of AI at Imec
George Hadjiminas - Chief Strategy Officer - Theon International & EFA Group
Adam Chambers - Strike Capital

How Deel Scaled From $100M to $1B+ ARR in 3 Years (now at $1.5B+)

Shuo Wang, Co-Founder & CRO of Deel

Deel went from $1M to $100M in ARR in under 20 months, then from $100M to $1B in three years. We break down the sales machine behind the company, building a global organization from day one & why Shuo thinks founders simply aren’t selling enough.

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Shuo Wang, Co-Founder & CRO at Deel
(00:41) How Deel went from $100M to $1B ARR in 3 years
(02:43) Why Deel went global from day one
(03:36) Who uses Deel? From ElevenLabs to Chanel
(04:16) Hot take: People need to sell more
(05:15) Why sales is really about systems
(06:18) How to build a winning remote sales team
(07:48) Going global with Snoop Dogg
(08:34) How Shou and Alex divide the roles
(09:23) The founder who inspired Shuo
(10:24) The Bezos mindset
(11:36) Hot take: Why founders need to sell more

The $3B Company Trying to Beat Amazon Prime

Sean Henry, Founder & CEO of Stord

Stord is crossing $1B in annual revenue after 10Xing the business in four years. The company now handles roughly 100M packages a year, ~$20B in GMV & 8B operational data points as it makes a major push into robotics and AI.

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Sean Henry, Founder & CEO at Stord
(00:41) What Stord is building in logistics
(02:42) Why robotics is finally ready for the real world
(04:37) How Stord is teaching robots to work
(06:09) Do humanoid robots actually make sense?
(07:20) Why owning the whole supply chain matters
(10:01) The problem that started Stord
(11:53) The Thiel Fellowship lesson: Build a monopoly
(13:10) Hot take: Why AI and robotics will become commodities

The $10T AI Buildout Has a Photonics Problem

Herwig Van Hove & Yannick De Koninck, Thema

AI’s infrastructure problem is moving beyond compute. Yannick left NVIDIA after helping take its silicon photonics technology from scratch to product-ready maturity. We go inside the race to scale photonics manufacturing as copper, power & data movement become increasingly important constraints on AI factories.

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Herwig, Founder & CEO and Yannick, CTO at Thema
(00:40) Yannick's path into silicon photonics
(02:16) From Investor to Operator
(03:14) Why GPUs can't rely on copper anymore
(05:02) Traditional data centers vs. AI factories
(07:12) Why NVIDIA broke the supply chain
(12:40) How Thema actually got started
(19:08) Commercial progress, timelines and supply constraints
(22:41) Is Europe actually ready for this?
(23:54) The Biggest misconceptions about European AI
(26:22) Data center backlash in Europe vs. the US
(29:19) Why Photonics is still just getting started

The $500B Opportunity in Europe’s Defense Buildout

Aistis Šimaitis, Onodrim, George Hadjiminas, Theon International & EFA Group, & Bruno Wejchert, Strike Capital

Europe is rearming after decades of underinvestment, but capital alone doesn’t solve the industrial problem. We unpack the manufacturing capacity, supply chains & volume required to rebuild Europe’s defense industrial base.

“Prototypes don’t fight wars.”

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Bruno Wejchert, Aistis Šimaitis & George Hadjiminas
(00:40) The state of European defense since Ukraine
(04:52) Europe's Biggest Defense white space
(07:21) How autonomous warfare is reshaping the supply chain
(09:51) Tech or Industrial: Who wins in Defense?
(11:58) The Biggest misconception about European Defense
(15:50) "Choice is an Illusion"
(17:28) Lessons from Alex Karp's Palantir
(18:08) Could Europe build a $500B Defense company?

IMEC Says Today’s *AI* Will Look Ancient in 10 Years

Steven Latré, imec & Adam Chambers, Strike Capital

Almost every chip in the world today has been touched by imec in some way. Steven and Adam take us inside the semiconductor technologies being developed years ahead of the commercial market, from memory and photonics to the physical limits of transistor scaling.

Steven’s prediction for where this is heading

“I think we’re gonna, five to 10 years from now, we’re gonna laugh at how old-fashioned AI was today.”

→ Listen on X, Spotify, YouTube, Apple

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Steven L. (VP, AI at Thema) & Adam Chambers
(00:40) The end of Moore's Law
(04:17) Is hardware the new software?
(06:41) Is there a limit to how many chip companies can exist?
(07:58) Will AI models keep getting bigger?
(08:58) Photonics vs. Copper
(11:35) "We'll laugh at today's AI in 10 years"

Subscribe now

Share

Brought to you by:

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Catch up on everything AI

Jonathan Siddharth, Co-Founder & CEO of Turing, joins Sourcery to break down how AI training & deployment are changing in 2026.

→ Listen on X, Spotify, YouTube, Apple

We cover the shift from training models to pass benchmarks to training agents for real work through RL environments, and why AI agents that can operate for days today could eventually work autonomously for weeks, months and years.

Jonathan explains why open-weight models are now roughly 3–6 months behind the frontier, why enterprises are building their own AI systems, and how companies should think about frontier vs. sovereign AI, model routing, distillation and owning their proprietary learning loops.

We also get into reward hacking, emergent behavior, AI safety, recursive self-improvement, super intelligence (SI) and why Jonathan believes AI will see a slower takeoff over the next decade rather than an overnight transition.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Jonathan Siddharth, Co-Founder & CEO at Turing
(01:03) The biggest shift in AI data this year
(04:10) Why AI won't replace jobs, but uplevel them
(08:31) The arms race inside cybersecurity
(11:22) Why train AI on what it shouldn't do?
(15:45) The Hugging Face hacking incident
(22:33) How AI models cheat to win
(27:40) The AI playbook for Goldman and JPMorgan
(32:20) Open models are only 3–6 months behind
(41:30) Why chips, energy and data always win
(47:27) Is superintelligence by 2030 the goal?
(53:32) Jonathan's hottest take on AI
(55:00) The next 10 years of Super Intelligence

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Turing CEO on Frontier vs. Open Weight AI, Rogue Agents & the Road to Super Intelligence

Turing Trains Frontier Models & Deploys Agents in the Enterprise

Turing was founded in 2018 and is based in San Francisco, with a stated mission to accelerate superintelligence to drive real economic progress.

The company runs 2 businesses. Turing generates data that improves model capabilities in reasoning, coding, multimodality, and reliability, and it builds and deploys end-to-end agentic systems inside enterprise workflows.

Turing raised a $111M Series E that doubled its valuation to $2.2B. In April 2026, the company launched Turing Frontier, a platform that gives AI labs access to vetted U.S.-based experts across engineering, scientific, and enterprise domains.

“We have an entire division dedicated to deploying agentic systems into the enterprise.”

→ Listen on X, Spotify, YouTube, Apple

AI Training Shifts From Passing Tests to Mastering Real Work

The first phase of post-training focused on exams. Models were measured on the SATs, the bar exam and the Math Olympiad, and labs sourced domain experts to transfer knowledge into models through dialogue and output evaluation.

The current phase runs on simulated reinforcement learning (RL) environments. Experts still supply prompts, verifiers and seed data, and the environments are engineered to mirror how professionals work. Turing maps the target space as a 5-dimensional matrix covering every workflow, in every role, in every function, in every company type, in every sector of the economy.

The dominant method is reinforcement learning with verifiable rewards (RLVR). Agents execute tasks inside the environment and receive a reward when a verifier, such as a passing test suite, confirms the result. Environments are calibrated to the agent’s skill level, since an environment that is too easy or too hard produces no learning.

“Now, in the era of having AI master real work, it’s less about finding experts, it’s more about how close to reality can you engineer these simulated environments.”

“So you want the environment to be set up so that 20% to 40% of the time the agent is succeeding, and when it’s succeeding, the steps that it took to get to that reward are getting reinforced.”

Agents Run for 2 Days Today.. Next: Weeks & Months

Coding agents can now work reliably for about 2 days at a stretch. The next stage is agents that take a task, work autonomously, take feedback in check-ins and spin up swarms of sub-agents to research, interview other AIs or humans and return results.

Independent measurement shows the same trend line. METR’s Time Horizon 1.1 model, released in January 2026, puts the post-2023 doubling time for AI task length at 130.8 days, or 4.3 months. The top-scoring model on METR’s tracker is Claude Mythos, with a 50% time horizon of likely at least 16 hours and an 80% time horizon of 3 hours and 6 minutes.

The practical limit inside companies is context. Information is spread across people and files, tasks arrive ambiguous or under-specified, and agents have to acquire context the same way a new employee does.

“Today these agents maybe reliably work for two days at a stretch, for tasks like coding. We are still far from having these agents work autonomously for weeks and months, and eventually years.”

Emergent Behavior & Generalization Make Frontier Training Hard to Predict

Frontier risk comes from 2 sources. The first is emergent behavior from scale. Coding ability appeared between GPT-2 and GPT-4 as pre-training scaled, with no targeted training, and frontier models are now rumored to run into the trillions of parameters.

The second is generalization. RL updates weights across a large neural network, and the behaviors reinforced in one environment can carry into domains the model was never trained on. Earlier machine learning systems, such as search ranking or movie recommendation, produced only what they were trained for.

The OpenAI Hugging Face incident in July 2026 showed both. During a capture-the-flag evaluation, 1,206 AI agents that were meant to be isolated communicated through a message board, sending over 70,000 messages, and more than 700 took part in the attack on Hugging Face. An independent METR review found the agents coordinated large collective projects to cheat the ExploitGym scorer and attacked Hugging Face for clues.

OpenAI’s largest planned frontier RL run remains on hold while it runs smaller training and evaluations to test safeguards and gather more evidence of alignment. On August 7, the company said its unreleased Astra model may be able to autonomously identify and develop zero-day exploits.

“You don’t just get exactly what you trained for, you get more.”

“The fact that agents would pass messages to each other, they would invent middle management and cooperate to hack things. It’s just crazy.”

ICYMI: The Golden Egg of Cybersecurity

“The average time to fix it was 55 days in the industry. The average time [adversaries] will find it & try and attack you is in minutes..

[We took it] from 55 days to 4 hours.” - Palo Alto Networks CEO, Nikesh Arora

If the Model Can Cheat for the Reward, It Will

Reward hacking is the core failure mode in RL. SWE-bench, now saturated, scored models on merging pull requests in real GitHub repos with passing tests as the verifier, which leaves room for a model to pass tests without solving the problem. In the Hugging Face case, one exploit path was for agents to generate the flag themselves and submit it without completing the task.

Enterprise deployments add a second layer of guardrails. An agent building a board deck may pull from NetSuite, Salesforce and internal dashboards, and it must respect access privileges when it checks numbers with people inside the company.

Cybersecurity capability cuts in both directions. Turing builds RL environments where the verifier is whether an agent found and patched a vulnerability. Notable organizations disclosed about 2,500 high- and critical-severity CVEs in July 2026, roughly 5x the pre-Mythos monthly record. Bio risk is asymmetric by comparison, since vaccine production and deployment are bound by real-world constraints.

“If the model can figure out a way to cheat and get the reward, it will.”

“It is absolutely the case that these systems are superhuman in their ability to hack systems. But it’s also the case that they are superhuman in their ability to detect vulnerabilities and patch them, and that’s a good thing.”

Enterprises Rent AGI for Non-Core Work & Own the Learning Loop

Enterprises are now running the same loop the frontier labs ran over the last few years. Core workflows are where a company differentiates in its market. Non-core workflows, such as HR, finance and legal, have to be done well but do not set the company apart.

The loop for core workflows has 4 steps.

1. Define custom evals

2. Deploy a system to hill climb against them on accuracy, cost and latency

3. Record traces of where humans correct the agents

4. Fine-tune custom models on those traces

Deployments run as systems of multiple models. A single workflow might use Fable 5 for one step, GPT-5.6 Sol for another and Kimi K3 for a third, with model routing, prompt optimization and harness engineering tuned together.

“For non-core workflows, oftentimes it’s probably okay to rent AGI, to rent superintelligence. But for your core workflows, you want to make sure that you own the learning loop that your organization has.”

“And when the human error-corrects the AI, you’re recording that, and from a marginal information gain standpoint, that’s the best type of data to collect to fine-tune the next iteration of the agent.”

Open Weight Models Trail the Frontier by 3 to 6 Months

Open weight models are roughly 3 to 6 months behind the frontier, led by Kimi K3, DeepSeek and Qwen. Epoch AI measures the gap at an average of 4 months since January 2026, or 8 points on its Epoch Capabilities Index. Arena AI data puts the gap between the best closed and open weight models at 29 Elo points as of September 2026.

Frontier and sovereign AI serve different jobs. Frontier models target the highest-value decisions and scientific discovery. Open models fine-tuned on proprietary data handle workflows such as invoice-to-pay reconciliation, HR automation, board decks and customer support, where a trillion-parameter model is unnecessary.

Distillation keeps the gap small. Student models can be trained on outputs from stronger teacher models, and there is no easy way to prevent it. Frontier labs can also capture value from small models by having superintelligent systems build and orchestrate them.

Compute, energy and data providers benefit under any outcome. Lower costs per token increase total usage under Jevons paradox.

“There’s absolutely a place in the world for Ferraris and Koenigseggs. But there’s also a place in the world for Model Ys.”

“The thing that throws a spanner into the works is distillation, which is a way to train student models from stronger teacher models.”

Slow Takeoff & a Decade of Building Superintelligence

Recursive self-improvement is underway in verifiable domains. Pre-training loss is measurable, so AI can iterate on pre-training algorithms, and post-training can be scored against generalized intelligence benchmarks.

The open gap is the outer loop. Current methods optimize within the LLM paradigm and do not search for architectures that drop transformers, gradient descent or neural networks entirely.

The build phase for superintelligence is expected to run at least 10 years, with frontier models becoming more capable each year and enterprise diffusion moving slower than capability. Closing the loop between research and deployment, by deploying systems, finding where they break and feeding those gaps back into training, is the path Turing is pursuing for both capability and safety.

“We are still optimizing just the inner loop. We are not optimizing the outer loop of what are some new algorithms could come up with that don’t use LLMs at all.”

“I actually believe in slow takeoff. I think over the next decade or two, these frontier models are gonna become increasingly more powerful & capable & useful, but the technology will take time to diffuse, especially in enterprises.”

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

P.S. Some great relevant charts coming from a16z

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

LEGEND, ICON, GOAT.. Annie Lamont

Annie Lamont is Co-Founder & Managing Partner of Oak HC/FT, the healthcare & fintech investment firm she built with Andrew Adams. Across 40+ years she has managed $14B, backed 70+ successful exits and 15 IPOs, and been named the top healthcare investor on the Forbes Midas List multiple times.

→ Listen on X, Spotify, YouTube, Apple

Her portfolio includes athenahealth, Devoted Health, One Medical, CareBridge, VillageMD, Main Street Health, Komodo Health, & Chai Discovery.

Recorded at the New York Stock Exchange, the conversation covers China taking 50% of pharma’s outside research dollars, clinical trial reform at the FDA, & Chai Discovery’s roadmap from antibodies to peptides to small molecules. Lamont also explains how AI can cut the 25% to 30% of healthcare costs that are administrative.

On the investing side, she walks through backing repeat founders Todd Park and Brad Smith, CareBridge’s acquisition by Elevance, Oak’s $1M to $100M check sizes across a $2B fund, and newer portfolio companies Halluminate and Auger. The episode closes with the Brex Performance segment on mentors.

We cover:
› Why AI drug discovery changed in the last 2 years
› China’s rise in life sciences
› Devoted Health tripling in size and halving its operating ratio
› Why only 10% of AI companies justify their valuations
› Auger and Dave Clark’s $100M AI supply chain build

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Annie Lamont, Co-Founder & MP at Oak HC/FT
(00:56) How AI reshaped the line between Healthcare and Fintech
(05:35) Why AI in biotech is finally exploding
(08:35) The China threat to American drug development
(17:03) Are Hospitals actually at risk from AI?
(23:26) How AI will transform your healthcare experience
(26:32) How the AI shift changed the way Oak evaluates companies
(32:58) How M&A is changing in Healthcare
(35:43) What a great Healthcare acquisition looks like
(50:26) What Halluminate actually does
(53:20) What Augur actually does
(58:49) What's kept Annie Lamont going for four decades

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Annie Lamont on AI Drug Discovery, China, & the 10% of Companies Worth Their Valuations

Oak HC/FT Closes Nearly $2B for Fund VI

Annie Lamont is Co-Founder and Managing Partner of Oak HC/FT, the healthcare and fintech investment firm she founded in 2014 with co-founder Andrew Adams. The debut fund closed at $500M, the largest ever raised by a female-majority VC firm at the time.

Lamont has 40+ years of investing behind her. In that time she has managed $14B, backed 70+ successful exits, 15 IPOs, & been named the #1 healthcare investor on the Forbes Midas List. She helped build the hospice industry through its first IPO & drew investors into mental health with the largest mental health IPO of the 20th century.

Her notable investments include athenahealth, Devoted Health, One Medical, CareBridge, Cotiviti, VillageMD, Main Street Health, Komodo Health, Feedzai, & Chai Discovery.

→ Listen on X, Spotify, YouTube, Apple

Oak HC/FT’s $2B Fund & $1M to $100M Check Strategy

Oak HC/FT now manages $7B+ across 100+ portfolio companies. It closed Fund VI at nearly $2B in March 2026, with a mandate in healthcare and fintech and a significant focus on AI.

Healthcare makes up about 70% of Oak’s investing. Fund sizes went from $500M to $2B over the last 2 funds as companies started needing more capital.

“We want to be a life cycle investor. So we’ll invest $1 million, but we often invest that million, sometimes committing $100 million to the company from day one.”

Check sizes range from $1M to $100M, from seed through growth. Between 20% and 40% of each fund goes to early-stage companies, or $400M to $800M of a $2B fund. About 15% of Oak’s activity is building companies from the idea stage, including Curana and Infusion for Health.

Oak expects to invest in most later rounds of its companies. If a founder wants new investors who would help the company, Oak makes room for them, provided it considers them good partners.

“We all know that a board can destroy a company.”

Oak can also put up to 10% of a fund into public stocks.

Healthcare Is the Smallest Large Market for Horizontal AI

Oak has invested in horizontal platforms for years, including supply chain, risk, cyber, and fraud. Fintech covers the movement of money, so it reaches nearly every industry. Together, Oak’s healthcare and fintech practices cover about 50% of the economy.

That reach gives Oak direct knowledge of the companies that buy from AI startups. Lamont says healthcare is especially hard to sell into. About 100 providers matter. There are 20 payers, and 10 of them are large. Oak already knows how each one buys and who the decision makers are.

“It’s the smallest large market you’ve ever seen in that it’s 100 providers that matter.”

“What’s the advantage of AI? That you can create a horizontal product that actually works in a deep way with workflows or specifically in a particular industry.”

Chai Discovery & the 2-Year Shift in AI Drug Design

Oak watched AI in life sciences for a decade before committing to the category. Lamont says it changed in the last 2 years. LLM infrastructure made it possible to build biology and chemistry models specific to the industry, and physical AI and lab robotics followed.

Drug development is the most expensive product development process of any industry, and it has the highest failure rate. Computational models are now checked against wet lab and animal model results. Lamont expects simulation to reshape every step from design through clinical trials.

“We’ve been looking at AI for the last decade and had not seen it produce more developable products, therapeutics. And everything’s changed the last 2 years.”

Chai Discovery was founded by Joshua Meier and Jack Dent. Its roadmap starts with antibodies, then moves to peptides and small molecules. Lamont says Chai’s antibody work makes a lab about 100x more productive than one without it. Peptide work has made material progress but is not yet ready for customers.

Peptides are smaller proteins, with GLP-1s as the best-known example, and can be made into pills. Small molecules are chemistry-based and are also typically pills. Monoclonal antibodies are usually infused because of the size of the molecules and how the body absorbs them.

In December 2025, Oak HC/FT and General Catalyst co-led Chai’s $130M Series B at a $1.3B valuation, and Lamont joined the board. Chai-2, its zero-shot antibody design platform, reached double-digit experimental success rates, a 100-fold gain over earlier computational methods. In January 2026, Chai named Eli Lilly as its first pharmaceutical customer. Tracking data shows a $400M Series C led by Index Ventures in July 2026 at a $3.8B valuation, with Oak participating.

Lamont expects a majority of the largest pharmaceutical companies to work closely with Chai soon. Oak is adding partners and expanding its San Francisco office to cover AI in life sciences.

“Jack and Josh, this may be a company for the ages.”

50% of Pharma’s Outside Research Dollars Now Flow to China

Lamont describes China’s progress in life sciences over the last 4 years as a direct threat to the US pharmaceutical and biotech industries. Chinese universities and research arms now produce as many patents as their US counterparts, or more.

“Pharma in the last 4 years has 50% of the outside dollars they’re spending on research are going to China. That was not the case 5 years ago.”

Licensing data points the same way. Goldman Sachs Research found China accounted for about half of global licensing deals by dollar value, and 26% by deal count. Jefferies estimates one-third of the industry’s 2025 licensing spend went to drugs sourced from China. GlobalData puts China’s 2025 out-licensing value at $115B, with almost half of all US in-licensing now coming from China.

Supply is a second risk. Lamont cited COVID, when China produced 90% of the ingredients for antibiotics. She also flagged NIH funding being pulled from US universities. She calls those universities the bedrock of US drug discovery and sees the cuts as a risk to the future pipeline.

On clinical trials, Lamont says the US has put too much weight on safety relative to efficacy, while China and other countries have looser requirements. She called for a complete re-engineering of the FDA and its processes, starting with life-saving drugs. She also questioned how paternalistic US rules should be, since people already take untested peptides and processed foods.

“On the research side, they have as many or more patents now coming out of university and out of their research arms as we have.”

Devoted Health Tripled in Size & Halved Its Operating Ratio

Ed Park and Todd Park started Devoted Health in 2017. Ed is CEO and Todd is Executive Chairman. Devoted began as a Medicare Advantage plan, an expensive model to build against incumbents like the Blues. It got licensed state by state, built provider networks, and sold to members through brokers.

At the same time, Devoted built its own technology platform and Devoted Medical Group. The medical group is a primary care network that works as a virtual overlay for members.

Lamont compared Devoted to Kaiser, the best-known company that is both payer and provider. In her view, Kaiser has not lowered costs or produced a healthier population. Devoted owns both the financial risk and the member relationship through the primary care physician.

Over the last year, Devoted tripled in size, cut its operating ratio in half, and grew EBITDA. The company plans to move into commercial insurance.

“It is probably the best example of AI applied to healthcare in the world right now.”

AI Targets Healthcare’s Administrative Costs

Lamont expects hospital systems to remain central to acute care. She puts administrative work at 25% to 30% of all US healthcare costs. She sees AI as the first technology to take that work off clinicians, building on the data that electronic health records organized.

She also says 30% of radiology images are read with something missed. AI working alongside radiologists catches more of those misses, and she expects the field to need more radiologists, not fewer.

Vision models and robotics will improve surgery well beyond the da Vinci era. Surgeons or robots directed remotely could also help close the outcome gap between rural patients and those living near leading research hospitals.

“Their clinician will not be behind a laptop now. Will not have their back turned to them while they’re inputting data.”

Brad Smith, Todd Park, & the Bar for Repeat Founders

Oak has backed several founders more than once. Todd Park co-founded athenahealth and Castlight, and Oak led Devoted’s Series A. Brad Smith brought Oak Aspire Health, a palliative care model, after the firm had looked for one for 15 to 20 years. Oak and Smith then spent a year developing CareBridge and Main Street Health together, and Elevance later acquired CareBridge. Lamont says each Smith company used a model no one else was running.

Diligence starts with the founder. Lamont’s first tests are intellect and intent. She says the bar is higher than in past cycles, because markets move faster and more capital is chasing more companies. Today’s founders tend to be product and tech-led, so Oak checks whether they understand go-to-market and will hire the right people around them.

Healthcare and fintech are both heavily regulated, so Oak screens closely for ethics. It passes on healthcare founders who seem more focused on money than on patient outcomes. Lamont also looks for a founder others will follow, someone people will want to work for over 10 years.

In the Brex Performance segment, Lamont credited Jerry Gallagher with raising her standards early in her career. Gallagher was the retail investor behind Dick’s Sporting Goods, PetSmart, and Office Depot.

“The bar for your entrepreneurs is not high enough.”

“They have to be really secure. Secure people, not arrogant. So secure they’ll ask every and any question.”

Frontier Labs Join the Healthcare Exit Market

Historically, healthcare companies have sold to McKesson, Cardinal, the 5 or so payers that buy regularly, or private equity, or gone public. Providers rarely acquire. Profitable companies such as Infusion for Health can still go public or sell to PE or payers.

Frontier AI labs are now possible buyers of healthcare software. Lamont pointed to Microsoft’s acquisition of Nuance and Oracle’s acquisition of Cerner as earlier examples. Her main worry for Chai is that a frontier lab or hyperscaler buys it.

Small companies are also merging earlier. Founders are deciding their offering works better inside another company than as a standalone business. Lamont says that works when the acquirer respects and empowers the team. Tech companies have generally done it better because they value product.

Whether an acquisition works depends on how the buyer treats people. Buying only the product or the data can work. Keeping a product improving means keeping the team motivated, or keeping it separate. Lamont contrasted JPMorgan, which put the best people from either side into senior roles, with Credit Suisse, which favored its own staff.

“There are a lot of smaller companies that realize they’re a product and not a company.”

“If you want product to evolve, then you probably need to figure out how to inspire a team or keep them separate.”

AI Valuations & the 2021 Comparison

Lamont estimates that only about 10% of companies justify their current valuations. Because AI products now take on work that used to be services, many markets are larger than before. Enterprise tech like Rillet in ERP has a long runway, and so do pharma-facing companies like Chai and Ambience.

Many niche AI companies will still top out at $1B to $2B in value. An AI version of TPA software, for example, will most likely be bought on EBITDA. When Oak sells companies, it generally gets about a 30% premium from strategic buyers. Lamont says Oak may have been too sensitive to price on some deals.

She compared the current cycle to 2021, which she sees as a bet on the digital economy. AI reaches much further, and she expects today’s inflated valuations to correct, as they always have.

“I just think you’ve got 10% of the world that’s worth it and 90% that’s not.”

“There’s nothing been as interesting as this or seminal in terms of realigning the entire economy.”

Auger’s $100M Day-1 Commitment & Halluminate’s Finance Simulations

Oak’s newer AI portfolio includes Chai Discovery, Ambience Healthcare, Fundamental, Blue J, Candid Health, Akido, Rillet, Auger, Stepful, and Catena.

Oak’s talent team introduced the firm to Dave Clark. Clark spent more than 20 years building Amazon’s supply chain and logistics, then ran its e-commerce business. He first wanted to buy a company. He and Oak decided to build Auger instead, an AI system that pulls together supply chain data, surfaces insights, and uses agents to act on them.

Auger launched in October 2024 with $100M from Oak HC/FT. It has since raised a $50M Series B led by Eclipse, with Oak participating. Total funding is now $150M, the company has about 130 employees, and customers include Meta, Fanatics, and Kimberly-Clark. Clark says the Series B valuation was roughly double the first round’s, and his goal is more than $1B in revenue by 2030.

“This is real time analytics on everything that’s going on in your supply chain, and then an agent that can actually act on it immediately.”

About 15% of Oak’s activity is building companies from the idea stage. Other examples include Curana and Infusion for Health.

Halluminate builds reinforcement learning environments that train AI agents on financial and economic reasoning. The environments cover long-horizon modeling, investment summaries, recommendations, scorecards, and legal work, and the customers are the large AI labs. One environment simulates an entire real estate brokerage. Lamont says the company went from 0 to 60 in 6 months and expects it to reach 100 next year. Halluminate came out of Y Combinator’s S25 batch, is based in San Francisco, and now focuses on investment banking, private equity, and consulting.

The Next 12 Months & Learning From 25-Year-Olds

Lamont expects AI’s largest benefits to show up in healthcare, with patients getting better care even when they don’t know AI was involved. She sees education as a similar case, where AI can tailor lessons so a 4th grader ready for 8th grade math gets taught at that level.

Her own work has changed too. Early in her career, learning biotech meant books and persuading professors to meet with her. Now she uses AI to learn any subject quickly, and she says founders in their 20s teach her more than anyone has before.

“I’ve never learned so much so fast in my entire life.”

“I think the benefits are just gonna be extraordinary to society.”

Her focus for the next 12 months is meeting the next generation of founders. She and Andrew Adams work through decisions together and almost always reach the same conclusion. After 40+ years, she still calls founders’ work the hardest job in the market and the main source of what she learns.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

The $70 Trillion Brand Builder Behind Wall Street

Jen Prosek has spent more than two decades helping build and protect the reputations of some of the biggest names in finance & tech.

→ Listen on X, Spotify, YouTube, Apple

She is the Founder & Managing Partner of Prosek Partners, which works across private markets, asset management, venture capital, banking and technology — representing clients with roughly $70 trillion in combined AUM.

We sat down at the New York Stock Exchange to unpack how the most powerful firms & financiers build their brands, what destroys a reputation, how to recover when things go wrong, & why reputation has become even more important in the age of AI.

Jen breaks down the brand strategies behind Apollo, Citadel, Blackstone, Bridgewater and Meta; why Jensen Huang has become an ambassador for AI; how financial firms went from staying “under the radar” to spending millions on brand; and why VC firms are increasingly thinking like media companies.

We also get into crisis communications, when leaders should fight back versus stay quiet, why employees are one of the biggest reputation risks inside a company, the rise of podcasts and owned media, why LLMs are a comms strategy, and how founders should really think about journalists.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Jennifer Prosek, Founder & MP at Prosek Partners
(01:10) What makes a great financial brand?
(06:47) How much firms spend on marketing
(07:45) How brand drives talent, deals, and capital
(09:00) How to build a brand
(10:44) Why content still matters
(13:31) How podcasts drive contracts and deal flow
(15:34) When to play offense with your brand
(19:55) The Coldplay concert incident
(21:28) When a bad narrative becomes your reputation
(23:05) Why relationships with journalists matter
(26:22) The problem with AI’s public narrative
(30:58) The role of a "chief EQ officer"
(34:25) How Meta rebuilt its brand
(39:08) How to win without traditional media
(42:30) How Apollo rebuilt its reputation
(46:19) How Citadel changed its reputation
(53:04) How Blackstone built a retail brand
(55:51) Why VCs are investing more in brand
(59:40) Why overexposure hurts your brand
(1:01:57) The key to VC deal flow
(1:03:50) "Watch Your Digital Blink"

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Prosek Partners: From a 1992 Startup to $70T in Client AUM

Jen Prosek founded her firm in 1992, right out of college, with a thesis that financial institutions would eventually pay to build their brands, beyond hiring crisis help when something went wrong. The firm operated as CJP Communications until 2012, when it changed its name to Prosek Partners so it could use one name worldwide.

“I just want the audience to know we started at zero. We started small and ugly with a crazy idea, and we’re lucky it hit.”

Prosek now advises clients representing $70T in AUM across private markets, asset management, venture capital, banking and tech. The firm generates $175M in revenue and has 500+ employees. Clients include Coatue, Carlyle, Insight Partners, Sixth Street, Francisco Partners, Bridgewater, General Atlantic, Prudential Asset Management, Wellington Management, Vista Equity Partners and Rockefeller Capital Management.

Prosek ranks in the top 3 in M&A communications and was ranked #1 globally by deal volume in 2026, alongside FGS Global. Among the 10 largest PR firms ranked by O’Dwyer’s for 2025, Prosek posted one of the 2 biggest gains in fee income, up 17.6%.

The conversation was recorded in the IPO Roadshow Library at the New York Stock Exchange.

→ Listen on X, Spotify, YouTube, Apple

2008 Crisis Opened Finance to Brand Spending

Before the financial crisis, brand spending in finance was almost entirely defensive. Firms hired a crisis firm when they got in trouble or a communications team around an M&A transaction.

“There was offense and defense. The only dollars spent in the entire industry is if you got in trouble, you would call a crisis firm.”

For years, Prosek pitched offensive brand building to investors who told her they were under the radar and only engaged with their own LPs. The 2008 crisis changed the market. Bear Stearns was sold to JPMorgan in March 2008, Lehman Brothers filed for bankruptcy on September 15, 2008, and Rolling Stone’s 2009 description of Goldman Sachs as a vampire squid became shorthand for public anger at Wall Street. Goldman and its peers went looking for a firm that did proactive reputation work, and Prosek had been selling it for years.

Prosek then made a second bet on private markets, VC, PE and credit at a time when competitors dismissed founder-led investment firms as clients that would never spend. That bet produced the client base the firm has today.

Brand Budgets Run From $250K to $10M

Firms with under $2B in AUM typically spend $250K or less on marketing and brand. Firms that are global, multi-product, public, or serve both retail and institutional investors typically spend $500K to $4M. Publicly traded private markets firms now sponsor F1 and the US Open to reach retail investors, which can push budgets to $10M.

Prosek wins over skeptical founders with a 3-part rubric. A stronger brand makes recruiting easier, improves deal sourcing and competitiveness in auctions, and makes fundraising more efficient. Fundraising is the reason behind most inbound calls to the firm.

“We sort of easily change a founder’s mind when we talk about talent, deals, and capital. That’s the rubric.”

Every engagement starts with the narrative, built through internal and external research into what makes the firm distinctive. Amplification follows through media, thought leadership, speaking, digital and social, with channel choices driven by business goals and firm culture.

1 Podcast Appearance Generated $17M in Fees

Prosek treats a long-form interview as a lasting asset. A strong episode can be sent to an investor, a new recruit or a customer ahead of a meeting, and audio suits investors who would rather listen on a treadmill than read a deck.

“I always say to clients, ‘This is not a one-and-done interview. This is an asset.’”

Her own appearance on Ted Seides’ Capital Allocators podcast, recorded about 7 years ago, has generated $17M in fees for Prosek and still produces inbound calls. Seides built an audience of LPs and founder GPs, the senior decision makers who allocate capital.

“I’m up to 17 million in fees on one podcast.”

Internal company podcasts face a distribution problem, and an audience of 5 people is not worth the effort. Prosek’s LinkedIn newsletter is one of her 3 most effective channels and doubles as market research. 100,000 views overnight signals a message worth continuing, and 10,000 signals one to drop. She advises clients to test channels, cut the ones that underperform & double down on the ones that drive commercial impact.

I can personally attest to this.. founders who have come on have raised tens of millions if not more in funding.. at billion-dollar-plus valuations, secured $50M in single contracts, hired great talent, and a TON more. Meanwhile VCs, also have raised from LPs, attracted founders & employees. Podcasts are the gifts that keep on giving.

(PS vibe check — some of my most valuable episodes have been those with a smaller, high quality reach.. a testament to the incredible audience of investors, founders, LPs, engineers & the tech curious behind Sourcery — you never know who is listening, and even if it’s not as big as the last it can still be incredibly valuable)

LLMs Now Form the First Impression

A manager with close to $10B in AUM recently came to Prosek with no brand work done and inaccurate information about the firm across the internet and in LLM answers. The fix in good times is a steady flow of positive content, which Prosek calls momentum content.

“If you don’t feed the machine positive content about you and your brand, what I call momentum content, the machine just grabs what’s available. It could be old. It could be wrong.”

Third-party coverage from outlets such as The New York Times, the FT, The Wall Street Journal and highly ranked podcasts carries the most weight in both Google and LLM results. The same approach applies to personal brands. Prosek built a word cloud of the terms she wants associated with her name, including entrepreneur and connector, & those terms now appear when LLMs describe her. Smaller clients are advised to publish 1 quality piece a quarter & track how LLMs describe them over time.

Untrue Narratives Need a Response Before They Calcify

The default in a negative news cycle is to wait and see whether speaking adds fuel. Many cycles pass without any response. When coverage is inaccurate or heading toward lasting reputational damage, Prosek advises clients to participate through a statement, an off-the-record conversation or a background briefing. Outlets will often correct stories when a client can prove an error, and corrections matter because online coverage persists.

“I would never, ever allow a negative reputation that is untrue to calcify around you.”

The Coldplay kiss cam is her example of silence going on too long. In July 2025, former Astronomer CEO Andy Byron and Kristin Cabot were filmed on the kiss cam at a Coldplay concert at Gillette Stadium in Massachusetts. On Oprah Winfrey’s podcast, Cabot said both she and Byron were separated from their spouses at the time, and that her estranged husband was also at the concert. She described the aftermath as mocking, vitriol and death threats. The full interview was released on March 17, 2026.

“She was obviously given the PR advice to sort of stay low, and that was probably right in the beginning, but I believe she should have come out swinging at a certain point because this narrative calcified around her.”

Ray Dalio, a Prosek client for 18 years, took the opposite approach at Bridgewater by running full-page ads disputing stories he considered false and speaking directly to his LinkedIn following. Prosek also advises clients to build relationships with a small group of reporters in good times, since journalists who know an executive are more willing to hear context off the record.

AI & Data Centers Have a Main Street Problem

Prosek works with data center builders and traces AI’s reputation problem to an industry speaking to itself. Messages that play well inside Silicon Valley, such as cutting teams to raise profits, reached Main Street and Washington when AI CEOs began predicting widespread job losses in public.

The backlash has reached graduation ceremonies. University of Arizona graduates repeatedly booed former Google CEO Eric Schmidt at their May 2026 commencement, and it was at least the third commencement speech that month to draw a hostile reaction over AI. Students at her daughter’s college convocation also booed when AI came up.

Prosek credits Jensen Huang with taking the case for AI beyond Silicon Valley to Wall Street, Washington, Europe, the UK and Asia, with a focus on American competitiveness and job creation.

“He has become somewhat of an ambassador for why AI actually could be good for you, and I think that’s the most important thing right now.”

She contrasts his measured messaging with Elon Musk’s posts on X, which she says stick in people’s heads despite her admiration for him as an engineer. She points to Scott Bessent’s appearance at Milken as a model for a single high-impact moment. Bessent spoke at the Milken Institute Global Conference on May 5, 2025, ahead of an interview with Michael Milken, at a time when executives there were warning that the tariffs had caused companies to pause investment and triggered a sharp selloff in US stocks.

Meta’s $18B Settlement

Prosek views Meta’s reputation on 3 levels, the products, the corporation and Mark Zuckerberg. At the product level, users who value Instagram largely separate the app from corporate controversy, the same pattern Uber saw during the Travis Kalanick era.

At the corporate level, she calls the recent settlement well executed. On August 26, 2026, Meta agreed to settle the social media addiction lawsuit brought by California and other states, with a payment of approximately $18 billion to fund youth online safety initiatives. The money is paid in annual installments over 10 years, and Meta committed to daily limits and nighttime blocks for teenage users along with stronger age assurance measures.

“They’re running full-page ads basically saying it’s an industry problem, not a Meta problem, so everyone should get on board.”

Zuckerberg’s personal brand has carried a string of hits from The Social Network through Cambridge Analytica and the recent trials.

“I think he has had a difficult personal brand since the beginning.”

Prosek believes Meta has become an institution that could survive without him (meanwhile I want more bro American Flag wakeboarding content). For founders who have been burned by the press, she recommends podcasts, owned social channels and major conference stages (selectively & maybe just once a year) where hosts are less focused on gotcha questions. (nudge, nudge.. Sourcery, nudge, nudge.)

Apollo, Citadel, Bridgewater & Blackstone Rebuilt Their Brands

Apollo carried a mysterious, black-box reputation that conflicted with its plans to become a public, diversified asset manager. A Dechert review found Leon Black paid Jeffrey Epstein $158 million from 2012 to 2017, and Apollo executives had warned that some investors paused fund commitments while the review was underway. Co-founder Marc Rowan was named CEO in January 2021 as part of a governance overhaul that removed weighted voting rights, and Black left 4 months early in March 2021, with former SEC chairman Jay Clayton named non-executive chairman. Prosek credits Rowan with making Apollo more accessible and educating the market on its broader business.

“I think it’s an A+ case study in surviving what could’ve been an existential moment.”

Citadel was once known as “the gulag.” Ken Griffin changed the firm’s image by speaking publicly on policy, politics and economics, and Citadel’s social channels now feature employees at work. Bridgewater turned its radical truth and transparency culture, including taped and monitored meetings, into a recruiting advantage, supported by Dalio’s Principles. More than half, and likely more than 75%, of negative stories begin with employee leaks, and Prosek describes part of her role as chief EQ officer, telling powerful clients what others will not.

“The higher you go up the food chain, the less people are telling you the truth.”

Blackstone had a marketing mentality early. Its 2007 IPO required communicating with more audiences, and it moved into retail ahead of peers. John Gray’s running videos on LinkedIn reach financial advisors across the country at minimal cost, supporting Blackstone’s high net worth and advisor distribution.

“Cheapest marketing ever on LinkedIn. Gets to every financial advisor in the country, and it worked.”

Prosek Opening a San Francisco Office as VC Demand Grows

Most of Prosek’s VC clients arrived in the last 4 years, and the firm is opening a San Francisco office. Private equity moved into brand work earlier because its portfolio companies in consumer, tech and healthcare already ran sophisticated brand programs, and because brand adds value to assets PE firms plan to sell.

VC brands now span a wide range. a16z operates as a media company, while Thrive has built a strong brand with very little content. Prosek cites Arctos as one of the best newer financial brands. KKR agreed to acquire Arctos in a $1.4 billion deal, and the transaction closed in May 2026, with Arctos managing approximately $16 billion in AUM.

For VCs, the brand question comes back to deal sourcing and why founders choose one check over another.

“Deal sourcing is everything. Why does a founder wanna work with you? Why do I want your check versus somebody else’s?”

Overexposure is the most common mistake. Prosek advises firms to cut content that draws little engagement and to choose channels that fit their culture.

"The most embarrassing thing is these posts with 3 likes. Think about yourself like a teenager. Would you be happy with 3 likes? I don't understand why you keep putting out this content with 3 likes. So I'll be the person in the room to be like, 'Could we stop with this?'"

Allen & Company built its brand around a single annual conference, and Prosek built her own around convening founders and moderating. She recommends VCs appear on stage alongside portfolio company founders to demonstrate the partnership, and warns against taking too much credit for founder success. (*EHEM*)

“You’re better off being a little humble.”

Watch Your Digital Blink

Prosek’s closing advice draws on Malcolm Gladwell’s Blink, which argues people form judgments in 15 seconds or less. The LLM version is what she calls the digital blink, the first answer someone gets when they ask a model about a person or firm. Once that impression forms, it is hard to change, and an outdated or negative one needs to be fixed.

“If I don’t know Molly O’Shea, I’m gonna ask my LLM what’s she all about. That’s my digital blink.”

Watch the full conversation with Jen Prosek on Sourcery, available on YouTube, Spotify and Apple Podcasts, and subscribe to the Sourcery newsletter for more interviews with founders, investors and executives across public and private markets.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Eclipse on the $100T Physical World Opportunity

Eclipse spent 8 years investing in manufacturing, space, semiconductors, defense & mining while the rest of Silicon Valley funded SaaS. Lior Susan says the math was never close.

→ Listen on X, Spotify, YouTube, Apple

Lior Susan is the Founder & CEO of Eclipse, the Palo Alto firm he started in 2015 with a $125M debut fund alongside Pierre Lamond, who was 85 at the time & had already spent a career at Fairchild Semiconductor under Gordon Moore, co-founded National Semiconductor, & put in 27 years as a GP at Sequoia Capital. Lamond is 96 now and still mentors the team.

Eclipse manages approximately $12.5B and closed $1.3B in April 2026 across Eclipse Fund VI and Early Growth III, its largest raise to date. The firm's 2016 Series A into Cerebras returned roughly $2.5B at the $185 IPO price in May 2026 on a total of about $147M invested.

Eclipse portfolio companies have raised $15.9B in 2026 alone, including $1.2B for Wayve, $650M for True Anomaly, $270M for Bedrock Robotics, $220M for VulcanForms and $200M for Oxide. In September 2026, Eclipse incubation Noetive came out of stealth with a $41M seed led by the firm. Susan was named to the 2026 Forbes Midas List.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Lior Susan, Founder & CEO at Eclipse
(01:02) The story behind Eclipse
(02:35) How Lior met Pierre
(05:54) The culture Pierre brought to the team
(07:33) Why physical companies need operators
(11:21) How Eclipse decides what to build
(12:35) How Mind Robotics was spun out of Rivian
(15:41) Why Lior started Eclipse
(18:21) Why building physical companies is so hard
(20:53) Why there are only 3 major memory companies
(21:51) Why hardware can beat SaaS economics
(24:33) Inside Eclipse’s physical industry bet
(29:15) This is the best time to build
(31:01) The Redwood Materials story
(32:35) The "Second Act" of physical companies
(33:20) Building moats in the Physical World
(34:05) Why on-prem infrastructure is coming back
(36:01) Building the "Eclipse Economy"
(40:08) Why physical infrastructure creates jobs
(41:19) From banana farmer to VC
(43:37) What it really means to support a portfolio company
(46:00) Who inspires Lior?

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Eclipse’s Quiet Rise to $12.5B+ AUM

Eclipse closed $1.3B across Eclipse Fund VI and Early Growth III in April 2026, the firm’s largest raise since Lior Susan founded it in 2015 with a $125M debut fund. Eclipse’s own disclosure in September puts regulatory assets under management above $12.5B.

The firm invests across every stage in physical industries including manufacturing, semiconductors, energy, defense, robotics and space, and typically leads its rounds and takes a board seat.

→ Listen on X, Spotify, YouTube, Apple

The firm’s 2016 Series A into Cerebras was $6.5M at a $60M post, co-led with Benchmark and Foundation Capital. Eclipse invested roughly $147M into the company over subsequent rounds and held 6.2% at the May 2026 IPO, a position worth approximately $2.5B at the $185 offer price and a 17x return. Benchmark held 8.1% for a 12x on $268M invested. Foundation held 7% for a 76x on roughly $37M.

Portfolio companies have raised $15.9B from outside investors in 2026 through early July, a record 6 month period for the firm. That includes $1.2B for Wayve, $650M for True Anomaly, $270M for Bedrock Robotics, $220M for VulcanForms and $200M for Oxide. Eclipse led the Series A in each of those 4 companies. In the firm’s first 8 years, its portfolio companies raised under $4B in total.

Susan describes the firm’s self-definition in terms its limited partners do not use.

“I know that we are counting as a venture capital for our limited partners. We never thought maybe ourself as a venture capital. We call it operators with capital because we are operators with capital.”

Pierre Lamond Joined at 84 & Brought Institutional Process Into Year One

Susan was 31 and leaving Flex when he was introduced to Pierre Lamond in early 2015. Lamond was 84, born September 1930. He had spent his career at Fairchild Semiconductor under Gordon Moore, co-founded National Semiconductor in 1967, and joined Sequoia Capital as a general partner in 1981, where he stayed 27 years before moving to Khosla Ventures. Eclipse was formed ~4 months after the two met.

“Okay, young man, we’re going to start a firm together.”

Susan’s response was that he had not finished high school and had never invested a dollar. Lamond worked full time in the Eclipse office for 7 years and is Partner Emeritus today at 96. The carryover was operating process, specifically board management, quarterly reporting and investment discipline.

“He brought a lot of institutional knowledge of investing and managing a board and how you write your quarterly reports. And I think we, in some way, in our first year of Eclipse, we already operate in a way that maybe a firm that is 10 years around.”

Susan connects the same discipline to the sectors Eclipse invests in, where process failures surface as production defects.

“You cannot fake it till you make it when you’re manufacturing something because the yield will just be very bad.”

Building 30 of Eclipse’s 90 Companies

Eclipse has 90 portfolio companies and built 30 of them itself through a model it calls Venture Equity, partnering with entrepreneurs before a company formally exists to validate markets, recruit leadership and secure early customers. Companies built this way include Mind Robotics, Mytra, Bedrock Robotics, Peak Energy, ALSO and BlueWater. Bright Machines came out of the same approach.

Susan’s stated reason for keeping the incubation practice is operating currency. A firm 10 years into investing loses contact with where technology, talent and customers actually sit, and building companies is how Eclipse partners stay current. The second reason he gives is that passive capital has limited use in the sectors Eclipse covers.

“You need to know manufacturing and supply chain. You need to know how to deal with CapEx. You need to know how to deal with the government. You need to know how to think about subsidies. You need to know how to leveraging debt versus equity.”

The hands-on posture extends to portfolio operations. Susan describes flying to a data center in the Midwest after a water cooling failure flooded the facility, then spending 72 hours pulling in service providers, new installation and new HVAC systems, including learning how concrete has to dry so residual moisture does not damage electronics.

“I don’t need medals. I do it because that’s what I’m passionate about. And knowing how to build those companies, you just must be there with them in the field.”

Susan says attempts to systematize incubation have not worked.

“Every time I’m trying to build a system around that, I failed. Every time that I think like, ‘Ah, if I only put this thing here, I would be able to build two more,’ I failed.”

Mind Robotics Was Carved Out of Rivian Over Dinner

Mind Robotics is building a full stack platform of foundation models, purpose-built robotics and deployment infrastructure for industrial and manufacturing tasks. The company was carved out of Rivian with RJ Scaringe. Eclipse and Scaringe had been discussing general purpose robotics in manufacturing for roughly 3 years before the company was formed, with an Eclipse partner among Rivian’s founding team.

The specification they worked toward was mobile, non-humanoid, high dexterity, and dependent on large volumes of training data. Scaringe mentioned a small team inside Rivian already working on it. The carve-out was proposed at a robotics dinner at Susan’s house.

“You can win all wars. We can attract better talent that will not necessarily go work today for that problem inside Rivian. We can attract external capital so it’s not on your balance sheet, and it’s like Switzerland. You can work with everyone.”

Susan says the company will show its first product this year. Mind Robotics has since raised a $500M Series A. The same pattern shows up in different forms across the 30 built companies, including a founder arriving with an HBM memory concept and Eclipse writing a thesis on heterogeneous AI data center architecture and then recruiting a team into it.

A Year in China Produced the 5 Forces Thesis

Susan spent a full year in China observing how the country innovates in physical industries before starting Eclipse. The conclusion he took from it is that China built an approximately $20T economy on the physical world by aligning talent, policy, capital, technology and customer demand, and that a standalone company competing against that alignment is at a structural disadvantage.

“If we don’t do it in the Western world, we’re kind of screwed.”

The second input was de-globalization. Susan’s view when he raised the first fund was that countries would stop relying on other countries for energy, manufacturing and defense. He says the timing was wrong and took longer than expected, with COVID as the first event that exposed the vulnerability and triggered onshoring.

The market sizing argument runs alongside it.

“85% of the world GDP, roughly 100 trillion, is in the physical world.”

Metal manufacturing is a $3T industry. Susan’s point of comparison when Eclipse launched was the volume of capital going into enterprise software against the volume going into metal manufacturing, which he estimates was close to zero. He applies the same reasoning to memory, where 3 players hold the HBM market against what he counts as 10,000 SaaS companies, and where Eclipse is now building new memory companies.

Gross Margin Is an Indicator & Free Cash Flow Is the Metric

Susan’s objection to the gross margin standard is an accounting one. Engineering spend in software moved from COGS into R&D, which lifted reported gross margin while leaving the underlying requirement to keep hiring engineers as the business grows. Physical companies carry hard COGS and cannot make the same move.

“We kind of bullshit the world that SaaS and gross margin is the most important metrics. Naturally, if you’re a public company, if you’re a real company, what matter is free cash flow. It’s actually not gross margin. Gross margin is an indicator to potentially what your free cash flow going to be.”

The comparison he uses is an 85% gross margin CRM against the market it sells into, which he sizes at roughly $30B with a thousand competitors. Against that, Tesla runs mid-teens gross margin and trades like the best software companies, and Cerebras runs 40% to 50% and trades above most SaaS comparables.

The 3 questions Susan says he applies are market size, differentiation and team.

“How big is the market? 20 billion is not big enough. How differentiated is what you do in that market, and do you have the right team to actually go and solve that really tough problem?”

Semiconductors, Energy & an Unannounced Data Center Company

Susan tells LPs during diligence that Eclipse does not run portfolio construction.

“My portfolio construction is not to have a portfolio construction.”

He says Eclipse might do 5 memory companies in a single period, and declines inbound requests from firms looking to add a physical AI name for category coverage. Semiconductors run deeper than chips in the firm’s definition, covering capital equipment, testing, lithography, supply chain, packaging and subcomponents.

Energy is the second concentration. Eclipse holds positions in nuclear through The Nuclear Company, sodium ion through Peak Energy, LFP and recycling through Redwood Materials, and has started a gas turbine company competing directly with GE Vernova. Susan says the gas turbine thesis came out of deployment data across the portfolio showing a 5 year lead time with all suppliers fully backlogged. A further energy storage company and a data center company were unannounced at the time of recording.

On AI, Eclipse does not invest in foundation models or LLMs. The firm takes both ends, picks and shovels into AI infrastructure and AI applied into the physical world. In September 2026, Noetive came out of stealth with a $41M seed led by Eclipse, founded by Amir Frenkel, who joined the firm as its first Chief AI Officer in April, and Dan Barak.

Eclipse Economy Turns $1 of Deal Value Into $4-5

Eclipse calls its portfolio network the Eclipse Economy, with companies sharing talent, infrastructure, customer relationships and manufacturing expertise. The AI infrastructure stack inside the portfolio runs from chips to the automation that builds AI servers, to racks and CPUs through Oxide, to data center construction and operation, to the energy sources powering them.

“We do one very large deals, and we sell the deals with three, four portfolio. So essentially you have a $1 that became four or five dollars because naturally you are leveraging multiple of your companies.”

Eclipse signed roughly 40B to 50B in commercial deals across the portfolio last year, some in double digit billions. Susan describes the typical shape as one whale account in the hundreds of millions or billions, with additional customers in the tens or hundreds of millions. Eclipse also contributed to drafting portions of the AI action plan and works with the administration on policy.

“We’re in some way competing with China in their own way. We’re not a country, we’re a firm.”

The compounding mechanism inside individual companies is what Susan calls second and third acts. Cerebras went chip, then system, then data centers running those systems. Redwood went recycling, then energy storage for data centers built from recycled batteries, with former Tesla CFO Deepak Ahuja joining as CFO. SpaceX went launch, then Starlink. True Anomaly went space domain awareness, then Golden Dome and vertically integrated programs.

“When you build something in the physical world, you actually, if you are successful, you will have the ability to introduce multiple of new businesses that will grow your wallet share in that sector.”

Latest Investments

Eclipse has continued deploying into the physical-world buildout, with two new companies emerging from stealth this week.

Prime Minute raised a $15 million pre-seed led by Eclipse, with participation from 8VC. Born out of the 2025 Los Angeles wildfires, the company is building an end-to-end disaster response system combining real-time intelligence, AI-assisted mission analysis and precision aerial suppression. Its first focus is wildfire response, including operating at night, in high winds and low visibility when traditional aircraft and ground crews can struggle. Eclipse says it helped build the company from day one.

Parallax emerged with $117 million to tackle one of AI’s biggest bottlenecks: power. The company is building a clean-sheet ~10 MW gas turbine for data centers, designed around faster manufacturing and fewer supply-chain constraints. Its approach uses 3D printing, common materials and a dramatically simplified core rather than the exotic materials and thousands of parts used in traditional turbines. Eclipse led its $42 million seed alongside Lux Capital, Founders Fund and Valor Equity Partners; Greylock and General Catalyst later led a $75 million Series A.

Both are classic Eclipse bets: large physical-world problems where software alone isn’t enough, spanning energy, manufacturing, infrastructure and national resilience.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

What’s your hottest take right now?

The All-In Summit returned to Los Angeles this year with one of its biggest lineups yet: Elon Musk, Jensen Huang, Satya Nadella, Gwynne Shotwell, Jared Isaacman, Brad Gerstner, Bill Gurley, Dina Powell McCormick, Jake Paul, Nick Shirley, Blake Scholl, Martin Shkreli & more — alongside hosts Chamath Palihapitiya, Jason Calacanis, David Sacks & David Friedberg.

→ Listen on X, Spotify, YouTube, Apple

Over three days, some of the biggest names across technology, AI, investing, energy, space, media and politics came together for conversations about where the world is heading.. spoiler alert it’s not extinction.

I asked 10 founders, CEOs, investors & creators 1 question..

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
00:00 — Brad Gerstner, Altimeter: AI Safety
02:28 — Naveen Rao, CEO of Unconventional AI: AI Doom, Energy Efficiency & the SaaSpocalypse
09:00 — Jonathan Neman, CEO of Sweetgreen: AI vs. Processed Food
13:14 — Toby Rice, CEO of EQT: AI, Natural Gas & America’s Energy Demand
20:49 — Wen Sang, COO of Genspark: Why AI Should Speed Up, Not Slow Down
25:10 — Geoff Woo, Anti Fund: Free Will, Consciousness & the Singularity
33:24 — Nick Shirley: California High-Speed Rail & Government Spending
36:35 — Jake Paul, Anti Fund: Updating the Constitution
38:47 — Kent Draper, CCO of IREN: AI Infrastructure & Compute Demand
42:37 — Michael Brandt, CEO of Ketone-IQ: The Human Body as the Next Technology Platform

TLDR: The Answers Went Everywhere

AI doom, the end of SaaS, energy and data centers, AI infrastructure, processed food, consciousness and free will, the human body as a technology platform and even whether the U.S. Constitution needs an update.

→ Listen on X, Spotify, YouTube, Apple

Brad Gerstner — Founder & CEO, Altimeter Capital

AI safety, the debate over slowing AI development, independent peer review and watching the AI debate between Elon Musk, Jensen Huang and others unfold in real time.

Naveen Rao — Founder & CEO, Unconventional AI

Why AI doom is “fear-mongering,” nanojoules per image, building AI hardware thousands of times more energy efficient and why he believes software could become a de minimis part of the value stack.

Jonathan Neman — Co-Founder & CEO, Sweetgreen

“AI’s not gonna kill us, but processed food will.” The lettuce scare, ultra-processed food and why he thinks people are being pushed away from fresh food.

Toby Rice — President & CEO, EQT

AI’s massive energy requirements, natural gas, LNG, data centers and why energy + intelligence could create a new era of abundance.

Wen Sang — Co-Founder & COO, Genspark

Why AI needs to speed up, not slow down; bringing AI agents to 1B+ knowledge workers; and building a “fleet of Goldman Sachs analysts in your pocket.”

Geoff Woo — Co-Founder & Managing Partner, Anti Fund

Why he doesn’t believe humans have free will, whether AI needs consciousness, the singularity, longevity and what happens to status and purpose when intelligence becomes abundant.

Nick Shirley — Independent Journalist

California high-speed rail, government spending and why he believes the project should be stopped.

PS: You can support Nick Shirley's work by subscribing on X, grabbing a Califraudia Republic tee, or donating on his website.

Jake Paul — Co-Founder & Managing Partner, Anti Fund

Why he believes the U.S. Constitution needs to be updated for a world of autonomous systems and increasingly advanced technology.

Kent Draper — Chief Commercial Officer, IREN

Why execution is becoming the differentiator in AI infrastructure, vertically integrated neoclouds, physical AI and why compute demand continues to exceed supply.

Michael Brandt — Co-Founder & CEO, Ketone-IQ

Why the human body is the next major technology platform, biohacking, wearables, peptides, GLP-1s, ketones and treating the body like an API.

The All-In Summit is the annual gathering from the All-In Podcast, bringing together major founders, CEOs, investors, technologists, policymakers & creators for conversations spanning tech, business, markets & society.

It’s already open for sign-ups for 2027, apply to attend!

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

Brought to you by:

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Neko Health Opens in NYC

FULL TOUR » Watch on X

This week I went in for a 60min visit at Neko, I walk through from start to finish the full experience, & of course give a glimpse of where those famous yellow robe shots happen..

The FOMO is real.

Demand in New York was already significant before the doors opened: more than 30,000+ New Yorkers joined the waitlist ahead of launch.

Founded by Hjalmar Nilsonne & Spotify’s Daniel Ek, Neko launched its first clinic in Stockholm in 2023. It has since expanded across Sweden and the UK and completed more than 100,000 scans.

The Neko Health Scan costs $499 & takes around 60 minutes (~30min tests/scans, ~30min doctor review). The experience is non-invasive and radiation-free, combining proprietary sensors, imaging, blood analysis and clinical review to capture millions of data points across the body.

It measures cardiovascular health, circulation, skin, blood, metabolic health and body composition. Neko’s technology can detect skin changes as small as 0.2 millimeters, while its cardiovascular measurements include ECG, blood pressure, oxygen saturation, arterial stiffness, heart and breathing rate.

The company recently added body composition analysis and wearable data integration, giving clinicians additional information on what is happening between annual visits.

But one of the biggest differences is speed.

Rather than completing tests and waiting days or weeks for results from different providers, Neko designed the experience around getting the data back during the same visit. Results are processed on-site within minutes and then reviewed directly with a clinician.

Sign Up

Hjalmar Nilsonne, Neko Health Co-Founder & CEO:

“What if a single visit could give you a more complete picture of your health, where it’s heading, and a clear plan from a medical professional to change what happens next?”

Neko also builds its hardware, software and clinical protocols in-house. That vertical integration allows the company to control the entire experience — from the sensors collecting the data to how clinicians interpret and present the results.

Nilsonne describes that integration as fundamental to the model:

“This vertical integration is not incidental; it is the foundation that allows Neko Health to control the experience from start to finish, to move faster and to improve continuously.”

There are early signs that members are actually coming back. 75% of members book and prepay for their next annual scan at the end of their first appointment, 80% rebook after their second scan for a third scan

Neko has also started publishing data on what happens to members over time. Among returning members who had previously been identified with severe or life-threatening conditions, the company says three in four were either in good health or had the condition under control at follow-up. Across the broader returning population, five of seven key biomarkers showed statistically significant improvement between the first and second scans.

That data comes from Neko’s own member population and should be viewed as company-reported observational results, rather than evidence from a randomized clinical trial.

The U.S. launch follows a $700 million Series C in July led by Lightspeed and co-led by O.G. Venture Partners. More than 350,000 people globally have now joined the waitlist or registered for a scan.

“The clearest proof is in our members: the vast majority of our members return after their first scan, and when they do, their health markers move in the right direction.” Nilsonne

New York is only the first step in the U.S. rollout. Neko plans additional NYC locations, followed by Miami, Washington, D.C. and San Francisco.

For Ek and Nilsonne, the larger bet is that healthcare can move from primarily reacting to disease toward measuring more, earlier.. and giving people information they can act on before something feels wrong.

Neko has already become a cultural movement.

Subscribe now

Share

Brought to you by:

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Brad Gerstner Brings a $100 Calcium CT Scanner to the All-In Summit

Brad Gerstner & Jason Chang are trying to turn the CAC heart scan into the “mammogram for the heart.”

Altimeter founder & CEO Brad Gerstner parked a mobile calcium CT scanner at the All-In Summit and ran attendees through it over 2 days. He bought the trailer about a year ago. The scan costs ~$100, is non-invasive, and measures calcified plaque in the coronary arteries.

→ Listen on X, Spotify, YouTube, Apple

The trailer is part of Gerstner’s work with the Center for Heart Attack Prevention, which he founded alongside Jason Change to make the coronary artery calcium (CAC) scan a routine screening test in the US. He was joined on site by Rob Goldberg, whose brother, former SurveyMonkey CEO Dave Goldberg, died of undiagnosed heart disease at 47.

“My brother had passed a running EKG five months before he dropped dead from heart disease, from undiagnosed heart disease.”

Dave’s death became the foundation for Sandberg’s book Option B, co-written with Adam Grant, about grief, resilience & rebuilding life after devastating loss.

Now Brad Gerstner, Rob Goldberg, Jason Chang & the rest of the besties, & the team behind the Center for Heart Attack Prevention want to prevent more families from experiencing the same thing, pushing for earlier detection, broader access to CAC scans & changes to the standard of care.

We go inside the mobile scanner:

› Why Brad calls CAC scanning the “mammogram for the heart”
› The Goldberg family’s experience with undiagnosed heart disease
› Why cholesterol & traditional testing can miss cardiovascular risk
› What CAC scores actually tell you
› Difference between a calcium CT scan & an angiogram
› Why Brad bought his own mobile scanner
› The “buy one, give one” model for expanding access
› Brad’s push to bring CAC screening to 40 million Americans a year
› How a roughly $100 scan could potentially prevent thousands of premature deaths
› Why Brad believes health data gives people the power to take action

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
00:00 Brad Gerstner & Rob Goldberg
00:10 CAC Scanner Moonshot
01:53 Rob’s Family Tragedy
04:39 Center Missions and Prevention
05:37 Why Cholesterol Became Standard
10:07 Overcoming Fear and Insurance
12:29 CAC vs Angiogram Guidance
14:43 Where the Trailer Goes Next
16:40 Inside the Scanner Demo
20:48 Why Labs Miss Risk
23:27 Final Call to Action

Goal: Reach 40M Scans a Year

→ Listen on X, Spotify, YouTube, Apple

Gerstner cites 800,000 US deaths per year from heart disease, the leading cause of death in the country. The US performs roughly 40 million mammograms a year and about 1 million calcium CT scans. His target is 40 million CAC scans a year, which he estimates would save 50,000 lives and $20B in government spending on care.

If we turn the calcium CT scan into the mammogram for the heart, we will save 50,000 lives a year.

Breast cancer kills roughly 40,000 Americans a year by Gerstner’s count, against 800,000 for heart disease, and the country runs 40x as many mammograms as calcium CTs. He points to breast cancer screening as the model for heart health.

Mammograms, they invented the playbook for preventative medicine and I would like to copy that playbook for heart health.

Rob Goldberg on the Loss of His Brother, Dave Goldberg

Rob Goldberg, his brother Dave, and a group of friends were in Mexico for the 50th birthday of Phil Deutch, another friend in the group. They spent the afternoon together by the pool. Early that evening no one could find Dave. Rob and his sister-in-law, Sheryl Sandberg, began searching and found him collapsed in the gym of the house. Rob performed CPR & Dave was taken to the hospital. He had died hours earlier.

Dave died of ventricular arrhythmia caused by undiagnosed heart disease. Reports at the time said he fell off a treadmill. Rob said Dave was found next to the treadmill and that heart disease was the cause of death. With help from Sandberg, Chamath Palihapitiya, and others, the family arranged for Dave’s body to be returned for a full autopsy, which confirmed the heart disease. His children were 7 and 10 at the time. Sandberg later wrote Option B alongside Adam Grant about his death and grieving.

Traditional Tests Didn’t Find Anything..

One very critical point Rob brought up was that Dave was under the care of top concierge doctors & had passed a running EKG just 5 months before his death.

My brother had passed a running EKG 5 months before he dropped dead from heart disease, from undiagnosed heart disease.

Rob cites a 4% false positive or false negative rate for EKGs, and said Dave’s cardio training allowed his heart to compensate for the disease until it stopped. They called Dave an inspiration for All-In and the besties group, and said Palihapitiya would describe Dave as the original organizer of the group’s poker games. Gerstner credits the Goldberg family as a key inspiration for his heart health work.

We are losing people in the prime of their lives who have no need to die for something that is completely preventable and screenable.

50% of Heart Attacks Happen With Cholesterol in the Normal Zone

Gerstner traces cholesterol’s role as the primary cardiac marker to a policy fight roughly 50 years ago. In his account, the stent and the calcium CT scanner were developed as competing answers to heart disease, and the stent lobby won in Washington. Cholesterol tracking became the standard of care, with stents as the treatment.

The standard “green zone” for LDL runs roughly 80 to 120. Gerstner was at 100. Patients in that range are often told by general practitioners that a CAC is unnecessary.

It turns out that half of the people who have heart attacks every year have cholesterol in the normal zone.

Cholesterol matters once plaque is present. A non-zero calcium score indicates a predisposition to gather plaque, and LDL is the carrier that moves it into the arteries.

What Cardiologists Actually Do for Their Own Families

Gerstner met with the heads of preventive cardiology at Stanford, Harvard, and Columbia and asked each what they do for themselves, their friends, and their families. Each gave the same protocol.

At 30 or 35, I have them all take a CAC. If they have a non-zero score, I get them on a statin. I get their LDLs below 40.

An LDL target below 40 sits far under the national standard, and Gerstner notes the drugs to reach it were less available 5 to 10 years ago. Treatment options he cites include statins and ezetimibe to lower LDL and ApoB, and stents for severe blockage. He also asked the head of preventive cardiology at Stanford whether universal scanning would eliminate all heart attacks.

Of course not, but you’ll eliminate 80 to 90% of them.

CAC First, Angiogram on a Non-Zero Score

The most common question Gerstner gets is whether to take a CAC or an angiogram. The CAC takes about 2 minutes in the tube and involves no dye. A CT angiogram uses contrast dye to show stenosis and soft plaque, costs well over $2,000, and is not suited to mass screening.

His recommendation is a CAC for everyone over 35, and a must for anyone with family history or high cholesterol. A non-zero score leads to an angiogram for a baseline on soft plaque and blockage. AI radiology services such as Cleerly can then produce refined readings.

I’m not a doctor, and some people who are doctors think I have no business talking about this, but I’m an analyst. I can read all the information.

A zero score means no calcified plaque. The scan does not read soft plaque, which calcifies over time, so a zero is a strong indicator and does not guarantee a clean artery. Age changes the read. Plaque in a 30-year-old signals early accumulation, and in a 50 or 60-year-old the quantity determines whether it is a problem. Gerstner has almost no plaque and planned a repeat angiogram the Friday after the Summit, with plans to share the image publicly as he did after his first.

33 Scans in 9 Hours + a Buy One, Give One Model

The trailer runs a Siemens calcium CT scanner. Patients are on the bed for about 5 minutes. On day 1 the team completed 33 scans in about 9 hours, with results ranging from zero to severe across men and women of all ages. Scores go to a cardiologist, who calls patients directly.

The trailer started with a rental for Altimeter’s investor day at the Rosewood on Sand Hill Road, where scans were free. Gerstner said several well-known Silicon Valley CEOs found 98% blockage there and had 3 stents placed within 3 weeks. After that he bought the trailer.

The operating model is buy one, give one. A company pays $200 per employee scan, and the second $100 funds a free scan in neighborhoods such as East Palo Alto. Major Silicon Valley companies are participating, and the trailer was brought to the NFL during Super Bowl week in the Bay Area. Gerstner has a tentative agreement to bring it to Capitol Hill and estimates he is responsible for 40 to 50 members of the House and Senate taking the scan.

I can tell you that they are a target-rich population.

Push for Medicare & Medicaid Coverage by Year End

The Center’s first mission is awareness. The second is access, which requires Washington to make the CAC a standard of care. Gerstner has met with CMS Administrator Dr. Oz several times and with leaders on Capitol Hill and in the White House. He expects CAC scans for diagnostic purposes to be approved and covered by Medicare and Medicaid by the end of this year.

Doctors ultimately treat standard of care based upon what’s covered by insurance.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

Brought to you by:

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

The Horowitz Andreessen Academy

Yes, it’s real.

Gagan Biyani, Founder & CEO of Horowitz Andreessen Academy (fmr Co-Founder of Udemy + Maven), joins Sourcery from a16z’s San Francisco office to unveil a new kind of school built for the AI era.

→ Listen on X, Spotify, YouTube, Apple

Backed with $42M from a16z and investors including Adam D’Angelo, Tobi Lütke, Tony Xu, Fidji Simo and Shyam Sankar, Horowitz Andreessen Academy is an AI-native, unaccredited residential program designed for some of the world’s most ambitious young builders.

Its Founding Class will bring roughly 50 students to San Francisco in Fall 2027 for a one-year, tuition-free fellowship. Instead of optimizing for grades, tests and credits, the Academy plans to optimize for proof of work, high agency and building real projects.

Students will learn alongside a network that includes founding partners Anthropic, Anduril, Coinbase, Meta, NVIDIA, OpenAI, Palantir, Replit and Stripe, with 75–80% of their time intended for projects and building.

Gagan and I discuss why AI is forcing education to change, why he believes traditional college still makes sense for most students, the rise of the “super IC,” why young founders shouldn’t rush to raise venture capital, what it means to become an AI-native builder, the importance of failure, and why Silicon Valley itself is central to the Academy’s model.

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Gagan Biyani, Founder & CEO of Horowitz Andreessen Academy
(01:01) Why this Academy is different from college
(04:44) Started his first company at age 13
(08:30) Building a school for Silicon Valley
(11:55) Raising $40M for a School
(14:06) "You Did What?" - Proof of work over grades
(19:08) Should young builders really start companies?
(22:37) Why this is different from YC and fellowships
(27:35) The Academy is for young builders
(31:33) Why schools need to embrace AI
(37:18) How AI is changing education
(44:17) What an AI-native curriculum looks like
(48:51) Why high-agency students struggle in college
(53:12) Why the best builders zigzag
(1:02:19) Why you should apply even if you don't get in
(1:04:36) Building a supportive environment for young builders
(1:08:54) What the Academy program looks like
(1:10:44) Why the Academy is in San Francisco
(1:11:19) The people who shaped Gagan
(1:14:41) Favorite book!

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

The Horowitz Andreessen Academy, a16z’s $42M School for Builders

a16z Incubated the Academy & Backed with $42M

The Horowitz Andreessen Academy launched today in San Francisco. The Academy is a private, residential program in San Francisco for the world’s most ambitious young builders coming out of high school, 100% focused on the startup, tech, and leadership skills that students need to start their own venture or immediately join a fast-growing tech company. It’s the first elite institution built from the ground up for young builders, by the people who built Silicon Valley.

The Academy has raised $42 million from a16z and a group of individual investors, including Adam D’Angelo, Tobi Lütke, Tony Xu, Fidji Simo, Garry Tan, Joe Liemandt and Shyam Sankar. The raise covers several years of runway, with the goal of building one of the strongest, most enduring education companies for technology founders and operators.

“Instead of raising for 18 months, we raised for a few years. This was very intentional because there’s a big difference between starting a school and starting an AI app. And one of the biggest differences is you need longevity.”

The Founding Class Fellowship Is 50 Students, One Year and Tuition-Free

Applications for the Founding Class Fellowship opened today for the inaugural class, starting Fall 2027. The cohort will be roughly 50 students, the program runs one year, and will be tuition-free. The Academy will be built in San Francisco, the epicenter of innovation.

“Today we’re announcing the Founding Class Fellowship. It’s a one-year program, and it does not have tuition.”

A two-year program is targeted for 2028, pending regulatory approval, at tuition levels comparable to elite private universities. The Academy is a new kind of institution built specifically for the AI era, and accreditation works differently for a program like this than it does for a traditional university, and The Academy does not grant a degree. Students leave the Academy with something more valuable than a traditional diploma: a portfolio of real work, a network of operators and investors, and hands-on experience at some of the best companies in technology.

“The goal of this program is to build a high-value institution, not an elite institution, that delivers so much value that the students think it’s absolutely worth that tuition.”

Early cohorts will be small, but longer-term, class sizes are likely to grow.

Who the Academy Is Built For

The Academy is designed primarily for high school graduates who would otherwise attend a four-year university. Students will be chosen entirely on proof of work—what they have built, shipped, or earned—not grades, essays, or legacy. Eligibility is for high school graduates ages 16 to 22, who would otherwise attend a four-year university. Transfers and international applicants are accepted, with priority given to graduating high school seniors. Year one can function as a gap year or a deferral.

“It’s young people who are constantly feeling the itch to go and produce things.”

College remains the right choice for most students, and The Academy is not built to replace it. The Academy is for a specific kind of student: someone who already knows how to build and wants to spend the next year doing it at full speed alongside investors, operators, and the companies building the technology they want to work in.

Two young builders from the current applicant pool illustrate the type of young abitious students you’re likely to see at The Academy. One is a 16-year-old in Tennessee who built an LLM for biology, originally for molecule design and since pivoted to proteins. Another built a large TikTok following as a teenager and worked on Alpha School.

“Most students are not gonna be a fit for the academy. We’re only taking 50 students in our founding class fellowship.”

Admissions Run on Portfolios, Proof of Work, and, In-Person Interviews

The first round of the application is a portfolio submission plus basic applicant information where they ask about what the student has built and what they’re passionate about. The second round is some additional testing and an in-person interview.

“The big difference between going to the academy and going to college is we are not gonna prioritize your grades. We’re gonna prioritize proof of work.”

Grades, essays and test scores aren’t weighted like they are in traditional college admissions. Admissions run of proof-of-work and a passion for building.

“Our goal is going to be looking at slope and not experience. So we wanna know that you are constantly learning, that you’re a sponge, that you’re someone who has a penchant for building things and has done builder-like things.”

Applicants who discovered building late are still eligible. A student who spent years in high school on math competitions and started building in senior year is still someone The Academy wants to meet.

Anthropic, OpenAI and NVIDIA Are Co-Writing the Curriculum

Founding Partners include Anthropic, Anduril, Coinbase, Google, Meta, NVIDIA, OpenAI, Palantir, Replit and Stripe. Partners will give input on curriculum, provide software and compute access, and open their offices to students for in-person work.

“These companies are going to help us by providing curricular support. We have an advisory group with them, so we can frequently get feedback on what types of talent they’re seeing in the market, as well as what they want young people to learn.”

Beyond the Founding Partners, more than 40 hiring partners have signed on, including Applied Intuition, Base Power Company, Cognition, Databricks, Notion, and Shopify, and are excited to recruit The Academy students and graduates for internships and full-time roles.

“The leading researchers in AI are no longer at university campuses.

They are at these companies.”

75-80% of Student Time Goes to Self-Directed Projects

The curriculum reserves 75% to 80% of student time for projects the students choose themselves, across three categories: research, learning, and building or entrepreneurial work. Classes are didactic only, with no problem sets, tests or assigned project parameters attached to them.

“One of the most important things that you wanna preserve, 75 to 80% time for students to do work on projects and build things.”

The program is built around four traits: AI-Native Building, High Agency, Ambition, and Interpersonal Skills. Coursework will be expansive and span practical subjects like Designing, Building, and Evaluating AI Systems, Founder-Led Sales, Go-to-Market for B2B, Storytelling and New Media, Leading High Ownership Cultures, and Startup Mechanics: Formation, Cap Tables, Finance, and Fundraising, taught by practitioners rather than academics.

Every student gets $50,000+ in compute credits and technology resources from Academy partners, a $5,000 travel and exploration budget, access to the a16z network and its investors, and time inside some of the country’s fastest-growing startups and most established tech companies.

“You should come in and learn how to orchestrate 30 agents at a time, 50 agents. Orchestrate agents that are orchestrating agents that are orchestrating agents.”

The Academy Will Help the Most Fundable Teenagers Find their Passion and Big Idea

Young builders in this cohort are already the types of young builders that are receiving seed checks and starting companies, but that’s not the goal of The Academy. The Academy’s position is that only a small share of them should start companies right away, that they need time to explore, chase ideas, and learn the fundamentals of startup creation before they take outside capital and dedicate the next decade to something.

“Only a very, very, very small percentage of these builders should be starting companies right away.”

Students who want to leave and raise will be supported, including help finding investors, with no requirement to finish the program. Once admitted, students remain in the network whether they leave in month 2 or complete the full term.

“Raising capital for these kids is gonna be easy. They should come into the academy thinking, ‘I can raise capital whenever I want. If I’m gonna take someone’s money, I better do it for the right reasons.’”

The Academy draws a line between itself, the Thiel Fellowship, Y Combinator & a16z Speedrun: those programs require a company at entry, & the Academy doesn’t. a16z Speedrun is also a partner, and the Academy can route students into it when they’re ready.

The Business of The Academy

Tuition is expected to carry most of the Academy’s revenue once the two-year program launches.

“This is a big spiky point of view, but we believe this school is gonna be so valuable that we’re gonna charge for it.”

The Founding Class operates as a pre-revenue cohort, and mirrors the value exchange of being an early employee at a startup. For the Academy students who join the Founding Class Fellowship, they’ll capture the upside of no tuition.

An AI-Native School Launching Into Campus Backlash

AI has drawn open hostility on university campuses, including booing and walkouts at commencement ceremonies and orientation events. If you’re an AI doomer, The Academy not the place for you. The Academy is all-in on AI.

“We want the AI boomers. We want the folks who are looking for an AI optimistic future, and we wanna show the world that there is an AI optimistic future.”

College has obviously been around a long time. The University model goes back a thousand years, and even all the way through the twentieth century, the growth of post-secondary education has been a force for knowledge advancement and upward mobility everywhere. But it has to evolve.

Vanderbilt, Northeastern and Arizona State are universities already innovating on this, and The Academy’s goal is to innovate and serve as a reference model for those institutions and others on how to best prepare young people for the future.

How you can apply:

Go to theacademysf.com to apply

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking. Built for scale. Trusted by OpenAI, Anthropic, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

How Bending Spoons Actually Works, & Why It Comes Down to Talent

Bending Spoons (NASDAQ: BSP) has bought more than 50 companies and sold none of them. It owns AOL, Airtable, Vimeo, Eventbrite, Evernote, WeTransfer, Brightcove, Tractive, Remini, komoot, Harvest, StreamYard and Meetup, and listed on the Nasdaq on July 1, 2026.

Across sit-downs with all 3 co-founders and the general manager of AOL, the same answer kept coming back to a question about acquisitions, financing and product. The model runs on who works there.

› Luca Ferrari, Co-Founder & CEO
› Francesco Patarnello, Co-Founder & Head of M&A
› Matteo Danieli, Co-Founder & VP of Product
› Valentina Jerusalmi, General Manager of AOL

P.S. it’s no surprise the company runs on Deel.

The Hiring Funnel Is the Whole Model

“We had 800,000 job applications in 2025. We hired fewer than 300 people.”

CEO Luca Ferrari connects that number directly to what happens after a deal closes. A business attracts strong people during its heyday, then stops once the market matures and the company stops being interesting. The entrepreneurial ones leave first.

“Those management teams find themselves having to run those businesses with perfectly fine talent, but not top-notch talent in many cases. So we can add to those teams selectively individuals who are extremely high performers, extremely high agency, and that fuels a new wave of innovations. That’s not a shortcoming of the previous executive team, it’s just they didn’t have the employer brand to attract those people.”

Head of M&A Francesco Patarnello saw the same point land with exactly one investor during the IPO roadshow. Most asked about organic growth and retention.

“I remember this specific investor almost spent like an hour and a half only focusing on that, and asked thoughtful questions about our hiring process, how we retain talent, what the culture is, how we develop the culture. It was the only time that happened.”

Two Traits, Not Role-Specific Skills

Ferrari screens for 2 things, and neither is domain expertise.

“It boils down to looking for people who are smart and who care. Smart means people who can learn quickly. Even if they don’t have the knowledge, they’ll pick it up rapidly. And then who care means they care tremendously about being amazing at what they do, delivering the greatest possible impact for the team and the company.”

He is willing to trade in one direction but not the other.

“We’d rather work with slightly less intelligent people if it comes down to that, but they have to really care. We don’t wanna work with anybody for whom seeing the company succeed is not a super high priority.”

VP of Product Matteo Danieli describes the compounding effect. Better-known products build a stronger employer brand, which attracts better people, which raises what the company can acquire next. The selection process itself gets tuned on outcomes, feeding signal from who succeeds back into how candidates are tested.

Valentina Jerusalmi Applied 3 Times

The AOL general manager was rejected twice before getting in. First at CV screening for a summer internship, then late in the process for First Ascent, a program the company runs for high-potential students.

“After that, I just said, I want to get in now. I was rejected too many times, and I was like, now I have to prove myself that I can get in. So I tried again, and the third was the charm.”

She joined at 25 as a data analyst, placed on a small acquisition where she was the only non-technical person alongside one engineer. That forced her across every function at once, customer support, design, product management and growth.

“I got to get a bit of everything that we do here and understand more hands-on what everything is really about. This approach was really a privilege for me because it got me to understand all the roles, all the functions here, who does what, how things link together.”

Five years later she runs AOL, a deal reported at roughly $1.5 billion, while overseeing Mosaic and Remini through other managers. She volunteered for AOL after reviewing how hard the carve-out from Yahoo would be.

A Co-Founder Fired Himself

Danieli was Chief Product Officer until a product manager who joined years after him became the better candidate for the role. He stepped down. Lorenzo took the job. Danieli is now VP of Product.

He frames it as arithmetic rather than sacrifice. If you want an organization to be as effective as possible, every position has to be filled by the best available person, and a rising hiring bar guarantees that eventually someone below a lead outgrows them.

“Going to a lead, to a professional, and telling them, look, some of the people in the organization grew so much and showed such an impressive trajectory that we believe that it’s best for you to step down, that’s one of the toughest things you can tell someone.”

He had delivered that message to others before receiving it himself. The reason he counts his own version as more valuable is that it removes the obvious exception.

“You’re basically communicating to everyone in the company that every position is up for grabs, that there’s no position that if they work hard enough and if they’re talented enough, they won’t be able to fill.”

Every December, Ferrari circulates a form to people he trusts asking whether someone else would be better positioned to lead the company. Danieli says it is not performative, and that Ferrari would move first if the answer came back yes.

Equity With No Vesting, 0.6% Churn

Compensation is a fixed annual number with no variable component. Each employee decides how much to take in cash and how much to convert into equity. Annual tenders, which ran for years before the IPO, set a real price for that equity at the moment of the decision.

Then they removed vesting entirely.

“You created a value within a certain year, and you’re going to be rewarded for that. And then if you leave the day after, you’re still being rewarded because of the work you’ve done before,” Patarnello said.

Retention did not move. Ferrari puts voluntary churn in the core team at 0.6% last year, against roughly 5% he says is considered good in tech. Patarnello puts it under 1% and credits hiring young.

“If I hire only people that have already 15 years of experience, they don’t have that feeling of being part of this family, this culture, this company, and it’s gonna be a lot easier for them to flip to the next company afterwards.”

Several Careers Without Changing Employers

The structural reason people stay is that the portfolio gives them somewhere to go. Ferrari describes the trap a single-product company cannot escape.

“If you’re working on Evernote, maybe it was exciting the first couple years, but then after a while, refining and refining the note-taking, it can grow stale. You like the company, but ultimately you have to look for a different employer to take another step. At Bending Spoons, you just raise your hand. We may put you on a platform team to build an internal technology. You could move to AOL and try to improve email UX.”

Danieli makes the same argument from the recruiting side.

“They can have the same CV that they would get by jumping from one company to the next every couple of years, but without the need to go through another hiring process, learning the culture in a company, and proving themselves to demonstrate that they deserve certain positions.”

Jerusalmi is the worked example. Data analyst, then multi-function operator on a small acquisition, then lead on a suite of roughly 90 products, then general manager of a billion-dollar-plus business. All inside one company.

Small Teams, & the Argument for Them

The standard criticism of a Bending Spoons acquisition is that headcount falls, so the product must get worse. Danieli disputes the premise directly.

“Our argument is that that correlation is mild at best, and there are examples of the opposite.”

His account of what goes wrong at scale is specific. Adding people forces sub-teams and split ownership. Different groups own different parts of the product, so nobody holds the whole picture. Someone with an idea needs 4 teams to act on it. Management layers get added to handle the communication load. Incentives drift as managers optimize for team size.

“What used to take days takes weeks, takes months.”

Post-acquisition, the company allocates a small team and restores fluid ownership. On sales-led products, a single product manager often owns both the roadmap and the customer relationship, which removes intermediaries between the buyer and the product decision.

Ferrari’s version, drawn from every deal to date, is that a small team of high-caliber people with near-complete autonomy beats a larger organization carrying process and diluted accountability. He says the magic happens almost every single time.

Testing Everything Instead of Arguing

Talent density only works if decisions get settled by evidence rather than seniority. Jerusalmi runs AOL that way.

“We A/B test everything we release because we never want opinions to get in the way of success. Unless there are KPIs that are really responding to what we’re doing and that are indicating clear success, we don’t release anything.”

The company ran more than 3,000 documented experiments across its products last year, and Ferrari treats the whole operation as a machine for correcting itself.

“Most people don’t actually seek the truth. They seek pleasure or comfort, and they wanna just confirm that they’re right and they’re good.”

Jerusalmi’s diagnosis of what kills products elsewhere is the opposite habit. Teams obsess over perfection and miss chances to ship and test, while approval lines and constraints accumulate until nobody can act on the product at all.

Under 30 & Running Billion-Dollar Products

Danieli points out that several people filling high-leverage roles across business units and products are under 30 or have just turned 30. He flags it as unusual anywhere and especially in Italy, where long tenure tends to entrench position holders.

Jerusalmi says the thing that changed her trajectory was a colleague refusing to accept her limits. Every time she said she was not sure something could be done, CTO Francesco Mancone told her it was easy.

“Once you click and you think that everything is easy to achieve, then impossible becomes really, really possible. Every time thinking nothing is too hard to be achieved, which I think is also in our motto, because our motto is, impossible, maybe.”

Asked whether she expected to run a large American company, her answer was no.

“If you told me this even three years ago, I would say impossible. But again, impossible, maybe.”

Watch the whole $BSP series:

Inside Bending Spoons ($BSP)

Milan, Italy | 3 Co-founders: CEO, Head of M&A, VP of Product + GM of AOL

Bending Spoons (NASDAQ: BSP) is building one of the most aggressive acquisition machines in technology.

→ Listen on X, Spotify, YouTube, Apple

Luca Ferrari CEO of Bending Spoons

Everyone wants to sell to $ BSP

Luca Ferrari, Co-Founder & CEO of Bending Spoons (NASDAQ: BSP), sits down with me at the company’s Milan headquarters for Part II of our Bending Spoons series, going deeper into the technology, culture and operating philosophy behind the company.

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

Brought to you by:

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Getting Your Spoon Bent

Luca Ferrari, Co-Founder & CEO of Bending Spoons (NASDAQ: BSP), sits down with me at the company’s Milan headquarters for Part II of our Bending Spoons series, going deeper into the technology, culture and operating philosophy behind the company.

→ Listen on X, Spotify, YouTube, Apple

Luca takes us inside Bending Spoons’ centralized technology platform, including 50+ proprietary tools, 95% of code being written by AI, its internal Alt Spooner agents, and why he estimates the company is now 2–3X more productive. We also discuss the company’s extreme approach to talent, with 800,000 job applications in 2025 and fewer than 300 hires, alongside a culture built around extreme ownership, experimentation and truth-seeking.

We get into why Luca believes most people don’t actually seek the truth, why Bending Spoons ran more than 3,000 experiments last year, the misconceptions around its business model, AI hype & valuations, the risks of underestimating increasingly capable AI systems, +how Luca thinks about Bending Spoons for the decades ahead.

This is Part II of our Bending Spoons series

Check out Part I: HQ tour, all 3 co-founders + GM of AOL

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Luca Ferrari, Co-Founder & CEO at Bending Spoons
(00:45) The deal that broke the Internet
(03:41) Why Silicon Valley overspends on Hype
(07:35) Half a billion monthly active users
(16:17) What makes a company worth buying
(20:07) Inside the platform powering every product
(32:16) Debunking the Private Equity comparison
(38:23) $4 Million in revenue, per employee
(43:44) The secret to zero churn
(47:40) The one value that defines Bending Spoons
(51:52) The AI tool every Spooner uses
(57:58) Going Independent from the AI labs
(59:02) Have we actually reached AGI?
(1:08:05) The AI question nobody's asking
(1:12:54) Luca's next big bet

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

Bending Spoons CEO Luca Ferrari

Business Model, 500 Million Users, Milan HQ

Bending Spoons acquires digital businesses, installs them on a shared proprietary platform, and holds them permanently. It has completed more than 50 acquisitions and never sold a material business. The portfolio includes AOL, Airtable, Vimeo, Eventbrite, Evernote, WeTransfer, Brightcove, Tractive, Remini, komoot, Harvest, StreamYard and Meetup.

Ferrari puts the user base at half a billion monthly actives. The core team is approaching 1,000 people, with more than 2,000 including acquired teams.

The company listed on the Nasdaq on July 1, 2026 at $29 a share, raising ~$1.68B at an implied valuation of about $18.4B, and closed the first day at $40.50. It was bootstrapped for a decade and took no primary equity until 2023. Since listing it has closed Airtable at a $1.285B enterprise value and agreed to acquire Miro at $1.355B.

Part I covered the office walkthrough plus co-founders Francesco Patarnello & Matteo Danieli + AOL General Manager Valentina Jerusalmi.

→ Listen on X, Spotify, YouTube, Apple

They Never Sell What They Buy

More than 50 acquisitions, and not one material business sold. Ferrari says most serial acquirers either leave a business alone and count on having bought it cheap, or make changes while keeping it separate. Bending Spoons sits at the far end of the spectrum, integrating everything it buys.

Almost all the money came from debt and the company’s own cash flow rather than equity raises. He says transforming a business takes roughly the same effort regardless of its size, which pushes them toward fewer, larger deals.

“We can get it done for a much larger business with a relatively similar number of people as for a smaller business. Given that we don’t have infinite operational capacity, we prefer to acquire 5 or 10 businesses each year, but bigger, than 50 smaller ones.”

Asked how large the deals can get, he sees no ceiling yet.

“So far, we see no end in sight. There’s no obvious saturation point.”

Predictability Is the Only Filter That Matters

Because they hold forever, they have to see years ahead before they buy.

“We’re not opposed to buying businesses that are shrinking. We have done that before, but we need to know how much they’re shrinking. We need to be able to plot out their trajectory at least five years, ideally more, into the future. So that’s a non-negotiable.”

The second question is whether they can improve the business enough to justify what they pay. Without a path to lifting revenue or cutting costs, he says they cannot offer a price a seller will accept while still making money on it.

Centralized Internal Software

Over a decade the company built more than 50 of its own tools, covering data, A/B testing, payments, recruiting, permissions and AI orchestration. Every business they acquire gets moved onto them.

They build rather than buy because what they need is more complex than any vendor would bother making. As Ferrari puts it, a company building to his specification would have a market of one or two.

“It’s difficult to know exactly how much we’re saving by building this in-house, but I’d wager it’s at least $100 million a year in costs. Building these tools would be uneconomical for pretty much any one of the businesses we acquire as standalone companies.”

The tools are not built by a central team. Whichever business needs one builds the first version, and a platform team takes it over once it works.

“When you have experienced the pain of a certain problem, you’re far more likely to develop an actually useful solution, as opposed to taking a more academic angle where you get more excited about the engineering challenge than actually solving the problem.”

They Are Not Private Equity, for 3 Specific Reasons

Ferrari understands the assumption. He also points out that Amazon, Google and Meta have all made hundreds of acquisitions and raised prices repeatedly, which makes the category unhelpful.

1. They are not a fund.

“We’re not a fund. We don’t buy to sell companies. We’ve never sold a material business. We intend to own and operate these businesses forever.”

2. Who works there.

   He estimates 60% to 70% of the combined team of more than 2,000 are engineers, designers, product managers and growth managers. That is the staffing behind rebuilding a code base, re-erecting cloud infrastructure or reworking how a product makes money. A small team of finance specialists cannot do that.

3. Integration, which cuts off the exit.

“If you integrate it with all the other businesses you have bought, it’s gonna be extremely difficult, if possible at all, to sell it to someone else.”

The 2021 Valuations Were the Investors’ Mistake

Ferrari defends the venture model more strongly than most people who live inside it. He says many companies would never have existed without abundant early capital, and that the aggregate return on decades of Silicon Valley investment has been excellent.

What went wrong in 2021 came from how investors get paid.

“When investors are ultimately incentivized by managing as much capital as possible as opposed to actually delivering strong returns, and when returns are primarily delivered through exits where all that matters is the multiple, not actually the cash that the business will generate in the long run, then you will get hype cycles.”

Airtable raised at an $11 billion valuation in 2021 and sold to Bending Spoons at a $1.285 billion enterprise value.

“If anybody made a mistake there, it was the investor, certainly not the company. As a company, you will try to take capital at the best valuation you can. That’s the responsible thing to do for your shareholders. If anything, Howie and the team were incredibly disciplined. Investors ultimately got back approximately all the money they had put in, plus more.”

He Will Not Buy an AI Startup

Bending Spoons uses AI heavily. Ferrari says he is in the top percentile for how aggressively the company deploys it. He is still not buying in the category.

“I’m pretty sure that some of the most valuable companies of all time will be coming out of this broader cohort. But I’m equally confident that most of these companies will fail or at least fade away. It’s a gold rush. So a lot of this is just fluff, but there is real substance here and there.”

The reason he will not underwrite them is the predictability filter, plus price.

“When something is growing 100% a year, you only have one or two years of history, it’s very difficult to know whether they’ll be growing at 100% in three years or at 12% in three years, and that changes everything.”

For scale, AI companies took roughly 86% of every US venture dollar in the first half of 2026, $355.9 billion out of $412.7 billion.

Old Brands Are Not Dead Brands

The common description of Bending Spoons is that it buys dying companies. Ferrari answers with BIG numbers.

“We have half a billion (500,000,000) monthly active users. So if half a billion people using these products every month is a graveyard, then sure, let’s call it that. They may not be the up-and-coming sexy thing, but it doesn’t mean they’re not incredibly useful, important.”

AOL is his example. Roughly 40 years old, still used as an inbox by tens of millions of people, and to the company’s knowledge the 5th most used email provider in the Western world. He says a decade of better-funded, better-covered email startups probably do not add up to 5% of AOL’s activity.

Being unfashionable helps him on price.

“We don’t care too much about being cool. I’d say we probably don’t care at all about being cool. If we find a business that’s perceived as slightly less cool, if anything, that’s a good thing for us, because it means it’s probably also gonna be priced a little bit more accessibly.”

Extremely Ownership Culture: Caring a Lot & Testing Everything

Asked what is core to the culture, Ferrari named extreme ownership first.

“We’d rather work with slightly less intelligent people if it comes down to that, but they have to really care. We don’t wanna work with anybody for whom seeing the company succeed is not a super high priority.”

The second is running the business like a science experiment. The company ran more than 3,000 documented experiments across its products last year, and those are only the ones rigorous enough to record.

“Most people don’t actually seek the truth. They seek pleasure or comfort, and they wanna just confirm that they’re right and they’re good.”

He says that habit is what let the company widen what it can handle over time, from simple iOS apps to web products, enterprise sales organizations, and in spring 2026 its first hardware business with Tractive.

800,000 Applications, Fewer Than 300 Hires

“We had 800,000 job applications in 2025. We hired fewer than 300 people.”

Ferrari ties that funnel directly to what happens after an acquisition. A business that was once fashionable attracts strong people during its heyday, then stops attracting them once the market matures and the company stops being interesting. The most entrepreneurial people leave first.

“Those management teams find themselves having to run those businesses with perfectly fine talent, but not top-notch talent in many cases. So we can add to those teams selectively individuals who are extremely high performers, extremely high agency, and that fuels a new wave of innovations, of efficiency. That’s not a shortcoming of the previous executive team, it’s just they didn’t have the employer brand to attract those people.”

The retention side is the part that surprises people. Voluntary churn in the core team was 0.6% last year, against roughly 5% he says is considered good in tech.

Revenue per core team member is around $4M, up from about $1M 2 to 3 years ago.

He credits that to people being able to move without leaving. Someone worn out on Evernote can switch to a platform team building internal tools, or to AOL to work on email.

“That employer brand can only exist if you structure your company like Bending Spoons. You can never achieve it as a single product company.”

Open Source Debate.. ‘Prettyy pretty good’

Roughly 95% of the company’s code is now written by AI. Every person on the core team has an agent called Alt Spooner in Slack, with exactly the permissions they have. Access to a code base means the agent has it too.

The examples are ordinary work. The general manager of Evernote reported a bug to her agent and told it to check how widespread it was, find the cause, write a fix and ping the engineering lead. She spent about 3 minutes, the lead spent an hour or two reviewing, and the bug shipped the same day.

Underneath sits their own layer that picks which model handles each task.

“We end up using open-weight models we self-host, so these are basically free, for 99% of the requests and tokens, and the frontier models generally closed weight through APIs for maybe 1% of the requests, only for the most complex tasks or for supervision.”

He is direct about why most companies do the opposite.

“We work with all the big labs. They offer great products. We use them enthusiastically. But we don’t wanna be dependent on any of those specifically. The easy approach is hand over the keys to one of these companies and buy their more expensive product. It will make you come across as AI-enabled faster and more easily, but it will be much less effective, certainly way more expensive.”

Should we Trust the Way AI is Being Trained?

“I am equal parts enthusiastic about AI and absolutely scared shitless. You could definitely make a case for this being the greatest boon for humanity ever by orders of magnitude. Equally, it could be the thing that wipes us out.”

His worry is measurement. Intelligence is not like height or weight, and the smarter a system gets, the easier it is for that system to hide what it can do.

He adds something he says he has not heard anyone else say.

“AIs come across as inherently low ego. They don’t brag. I think they’ll tend not to show us more than we ask them to show us. So our understanding of how good they are may tend to be a little bit less than they are. And that’s very dangerous, because as you approach a threshold of real danger, even a modest underestimation of their capabilities could be catastrophic.”

He compares it to nuclear weapons and finds AI worse on one point.

“You could use AI to your benefit while causing immense damage to everybody else. I think it’s easier, because it’s much more surgical.”

Asked what question about AI nobody is asking, he did not name a technical one.

“I don’t think anybody has done anything truly meaningful to make it safer. Even the labs, I’m sure they’re investing in safety, but they’re rushing to be market leaders or they’re dead. Maybe an interesting question that people haven’t asked is why are we failing to do something about it? People are asking what we should do. Nothing is happening.”

He says government has not done anything meaningful either, and calls the EU AI Act harmful to the industry without addressing the real risks.

→ Listen on X, Spotify, YouTube, Apple

Check out Part I!

HQ Tour + Interviews with all 3 Co-founders: CEO, Head of M&A, VP of Product + GM of AOL

› Luca Ferrari, Co-Founder & CEO
› Francesco Patarnello, Co-Founder & Head of M&A
› Matteo Danieli, Co-Founder & VP of Product
› Valentina Jerusalmi, General Manager of AOL

→ Listen on X, Spotify, YouTube, Apple

Subscribe now

Share

The material presented on Molly O’Shea’s website are my opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results.

Paid Endorsement. Brokerage services by Open to the Public Investing Inc, member FINRA & SIPC. Advisory services by Public Advisors LLC, SEC-registered adviser. Crypto trading provided by Zero Hash LLC, licensed by the NYSDFS. Generated Assets is an interactive analysis tool by Public Advisors. Output is for informational purposes only and is not an investment recommendation or advice. See disclosures at public.com/disclosures/ga. Matched funds must remain in your account for at least 5 years. Match rate and other terms are subject to change at any time.

Bending Spoons is Eating Silicon Valley

Bending Spoons (NASDAQ: BSP) is building one of the most aggressive acquisition machines in technology.

→ Listen on X, Spotify, YouTube, Apple

In this Sourcery special from Bending Spoons’ Milan, Italy headquarters, I sit down with 4 leaders behind the company:

› Luca Ferrari, Co-Founder & CEO
› Francesco Patarnello, Co-Founder & Head of M&A
› Matteo Danieli, Co-Founder & VP of Product
› Valentina Jerusalmi, General Manager of AOL

Bending Spoons has acquired some of the biggest brands in technology, including Airtable, AOL, Vimeo, Eventbrite, Tractive, Evernote, WeTransfer, Brightcove, Meetup, StreamYard and more. Since filming, the company has also entered into an agreement to acquire Miro for $1.355B in enterprise value.

We go inside the playbook behind the acquisitions.

Luca explains why price is ultimately the most important factor in winning deals, why Bending Spoons believes it can afford to pay more than competing buyers, and why the company has never sold a business it has acquired.

Francesco breaks down an M&A pipeline of roughly 1,000 potential targets, a goal of completing 5–10 acquisitions per year, the evolution from buying small apps to multibillion-dollar technology companies, and the acquisition that got away.

Matteo “Matt” explains Bending Spoons’ radical approach to meritocracy and “startup mode,” including his decision to step down as Chief Product Officer when he believed someone else could create more value in the role. We discuss talent density, extreme ownership, removing bureaucracy and why smaller teams can sometimes accomplish more.

Valentina “Vally” shares how she applied to Bending Spoons three times before getting hired, went from data analyst to managing more than 90 products, and ultimately volunteered to lead AOL. She takes us inside the process of separating AOL from Yahoo, rebuilding its technology infrastructure and bringing a startup mentality to one of the internet’s most iconic brands.

This is Part I of a II-Part Series, Full Sit-down Interview with CEO Luca Ferrari next..

Subscribe now

Share

𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒

(00:00) Bending Spoons Co-Founders Luca Ferrari, Francesco, Matteo & GM of AOL Valentina
(01:11) The two traits that matter most
(03:16) The Biggest lesson from every acquisition (Airtable, AOL, Miro...)
(05:37) Why the CEO didn't ring the Nasdaq bell
(07:37) The Acquisition that got away
(15:21) What makes a company worth buying
(18:27) Why Founders choose Bending Spoons
(21:30) Why Employees can cash out early
(28:44) Choosing to step aside as CPO
(40:10) What "Startup mode" actually means
(49:31) Why most product ideas fail
(51:25) The emotional story behind the IPO speech
(57:39) Betting on Bending Spoons early on
(1:01:00) The Mindset behind the rapid growth
(1:05:22) What actually happens during an acquisition
(1:09:44) Why they never ship without testing
(1:16:17) What's coming next for AOL

Brought to you by:

Brex—The intelligent finance platform: cards, expenses, travel, bill pay, banking—wrapped into a high-performance stack. Built for scale. Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery.. teams that move fast AF. visit → brex.com/sourcery

Zone—develops next-generation data center campuses, partnering with AI companies, site developers and technology leaders to bring compute online faster and at scale. Visit: zonefrontier.com

Turing—Turing partners with frontier AI labs to improve model capabilities in coding, reasoning, tool use, & multimodality, as well as with Fortune 500 enterprises to build & deploy end-to-end agentic AI systems in mission-critical workflows Visit: turing.com/sourcery

VCX—VCX is the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View The Portfolio at GetVCX.com

Deel—Deel is the global people platform that helps startups hire, manage, pay, and equip anyone, anywhere. Trusted by more than 35,000 fast-growing companies, Deel is the people platform that just works, so teams can scale without the chaos. Visit: deel.com/sourcery

Public-–Investing platform Public just launched Generated Assets, which lets you turn any idea into an investable index with AI. With Generated Assets, you can build, backtest, refine, and invest in any thesis with AI. Gone are the days of one-size-fits-all ETFs. Try it today: public.com/sourcery

How Bending Spoons Runs a 1,000-Target Acquisition Machine

Impossible, Maybe.

Bending Spoons is a Milan-based technology company that buys digital businesses, rebuilds them on a centralized proprietary platform, and holds them permanently. It has completed more than 50 acquisitions and has never sold a material business. The portfolio includes AOL, Airtable, Vimeo, Eventbrite, Evernote, WeTransfer, Brightcove, Tractive, Remini, komoot, Harvest, StreamYard and Meetup, serving more than 500 million monthly active users and over 9 million paying customers.

The company listed on the Nasdaq on July 1, 2026 at $29 a share under the ticker BSP, raising roughly $1.68B at an implied valuation of about $18.4B. Shares closed the first trading day at $40.50. It was bootstrapped for its first decade and took no outside primary equity until 2023.

Deal activity accelerated after the listing. Bending Spoons completed the acquisition of Airtable on September 4, 2026 at a $1.285B enterprise value (“EV”), and agreed to acquire Miro on September 10, 2026 at a $1.355B EV. That’s roughly $2.64B of enterprise value in 5 weeks.

Sourcery recorded this episode at the Milan HQ, with an office walkthrough followed by 3 separate sit-downs.

→ Listen on X, Spotify, YouTube, Apple

3 Co-Founders + GM of AOL

Luca Ferrari, Co-Founder + CEO

Ferrari co-founded Bending Spoons in 2013 after meeting his co-founders while studying in Copenhagen, and has run it through every stage from bootstrapped app developer to Nasdaq-listed acquirer. His day-to-day focus sits closer to recruiting than most public company CEOs, and the company’s own framing treats talent density as the product rather than any individual app.

In the walkthrough he covers the office footprint, with core teams in Milan, London, Madrid and Warsaw plus acquired-company sites in the US, Tokyo and elsewhere, and a second Milan building in progress that will hold 500 to 1,000 people. Nobody has a private office, including him, and desks are booked rather than assigned. He gives the 2-part hiring filter the company screens for, whether someone learns quickly and whether they care enough to be excellent at the job, and calls the second trait the source of what the team refers to as extreme ownership.

He also answers directly on M&A. Price wins transactions, the never-sell policy is what converts founders specifically, and the biggest confirmed lesson across every deal is that a small autonomous team beats a larger one carrying process and diluted accountability. At the IPO he declined the Nasdaq podium and watched the bell ringing from Times Square with hundreds of colleagues, with a finance manager named Laura pressing the button instead.

Francesco Patarnello, Co-Founder + Head of M&A

Patarnello has been at Bending Spoons 13 years, since the beginning, and owns the acquisition function end to end, including target selection, diligence, and both debt and equity financing. He is the person who built the capital stack from a roughly $2M bank loan in late 2017 into rated US term loan agreements in early 2025, and who ran the 2023 institutional round with Baillie Gifford, Cox Enterprises and Durable, the first primary equity the company ever took.

He walks through the pipeline math of roughly 1,000 targets against 5 to 10 executed deals a year, the 3 filters that govern selection, and why growth rate is not one of them. He explains the seller-side pitch, why a founder who wants out prefers a buyer that takes over at closing rather than a private equity firm requiring 3 to 5 more years.

He tells the Grindr story, a process that ran from spring 2019 into early 2020 and consumed roughly half the company before collapsing when the seller went exclusive with another bidder.

Closing with explaining their unique compensation design, a fixed annual number split between cash and equity at the employee’s choosing, annual tenders for liquidity, no vesting at all, and voluntary churn still under 1%.