Brendan Foody

1. Key Themes

Humans move from doing repetitive work to teaching agents how to automate it

Foody's trillion-dollar thesis starts by working backward from how humans contribute to the economy. Mercor's whole business is the infrastructure for that shift. Brendan Foody (CEO, Mercor): "Right now most work is relatively repetitive, where people will repetitively respond to customer support tickets or repetitively red line a legal contract... when in the future it's going to much more look like training an agent one time how to automate these repetitive tasks." 00:01:01 He sees enterprises already moving toward "people contributing to the economy through creating evals, updating skills files, training agents how to automate all this repetitive work." 00:01:01

Mercor is not a marketplace: the product is the RL environment, and experts are one ingredient

Foody calls this the largest misconception about the business. Brendan Foody: "People think about us as a marketplace like Uber... when actually the experts are one component of the datasets, the RL environments, that we deliver to the labs." 00:05:51 An RL environment has three parts: apps (clones of Google Workspace and similar tools), worlds (messages, slides, docs, sheets), and tasks (prompts and rubrics). "The barrier to automating the entire economy is how do we cover the full distribution of all of the RL environments that mirror everything in the distribution economy." 00:05:51 Mercor runs three platforms behind this: a talent network (5M+ people seeking work), Studio (a data platform with an AI copilot and agentic QC), and an AI project manager that autonomously runs projects, hires, and is "accountable to the customer outcomes." 00:02:42

Jevons paradox in expert labor: productivity gains expanded demand instead of shrinking it

Agents make each expert hour more productive, and that grew the market. Brendan Foody: "An hour of their time might get 3 to 5x that because they're able to use an AI copilot... what's happened, as we've made the experts more productive, is that it's been this case of Jevons paradox, where the demand just expanded exponentially because ultimately like 99% of knowledge still lives in people's heads." 00:07:35 The scale of the ramp: "If we're paying out call $5 million a day, as of 2 years ago, 24 months ago, we were paying out a million dollars a year." 00:14:46

Models need humans at the frontier because self-grading doesn't work

Foody answers the "non-durable category" objection by pointing to how learning works. Brendan Foody: "The market for humans teaching models how to do something is bound by the amount of things that humans can do that models can't do across either the digital world or the physical world... we need humans that measure that frontier." 00:09:12 His analogy is a student grading their own essays. "It's like pretty difficult to measure what is beyond the frontier of your own capabilities and that's why we have feedback." 00:09:12 Domains like math have a calculator-style check, but most of the economy doesn't.

Every Fortune 500 becomes a "neo-lab," and inference spend is the trigger

Foody lays out a concrete adoption heuristic keyed to inference bills. Brendan Foody: "When a company is spending under $10 million on inference, they don't generally care that much about evals... Once they go from the 10 to $100 million on inference range... roughly 10% of their inference budget should be going to eval... Once they're past $100 million in inference, like a Harvey or a Cursor... they need to be really seriously thinking about how do they post-train models, because the economics start to check out." 00:44:46 He adds that "every Fortune 500 is going to look like a neo-lab, where they have their core datasets, their core distribution that they care about." 00:43:45

AI data sovereignty: enterprises won't hand their competitive advantage to the labs

Foody pushes back on the narrative that labs will train on enterprise usage data. Brendan Foody: "The reason that a company like Goldman Sachs exists is that they have a set of competitive advantages associated with the people, the processes, the documents... and turning on allowing a frontier lab to train on that data would be analogous to just sharing all of it with a competitor." 00:44:46 He expects Anthropic and OpenAI to be enormous ("both of them could very well be $10 trillion plus companies") but argues value must accrue to the rest of the Fortune 500 through their own models.

Where moats live in the AI stack: brand and data at the app layer, custom models at the inference layer

Foody gives a layer-by-layer moat analysis. At the app layer: "The two things where companies are able to build moats that I've seen that have been durable, is, first of all, their brand... and then second of all, their data moats." 00:37:05 At the inference layer: "The switching costs of reselling GPUs are zero... the real switching costs and durable margins come from how do you help a customer customize a model that has high switching costs." 00:38:46 He argues the app-layer survivors "need to be building the frontier model in their specific domain, like Harvey building the frontier model in law, because ultimately the model is the product." 00:35:20

How to train models: build tasks and start with the eval

His advice to enterprises on post-training is concrete. Brendan Foody: "The vast majority of where the compute gets spent on in training is no longer just like read all the documents in the company... Most of the compute is now going towards RL, where the model will attempt the problem many times and then get a score on one of these rubrics." 00:48:25 The two pieces of advice: "build tasks rather than just like pretending the model can read everything and learn, and... they should start with the eval." 00:48:25

Culture as a competitive weapon: high standards, time travelers, customer obsession

Foody names three operationalized values behind winning in a crowded category. Brendan Foody: "High standards, just having like a really high hiring bar... time travelers, which is about being super like long-term oriented... and then the third thing is customer obsession. Like we are an incredibly hardworking customer-obsessed company where we move mountains for our customers." 00:16:25 He also bans vanity metrics: when an exec posted a weekly revenue graph, "I was like, 'No, I don't want the company like looking at this'... the company should be looking at how happy our customers, how happy are our experts, how well utilized our experts." 00:22:00

2. Contrarian Perspectives

Data is still contrarian, and the consensus has been wrong for three years

Investors assumed superintelligence would arrive within about 3 years and eliminate the need for human-generated training data. Brendan Foody: "There has been this perception by investors that we will get superintelligence in 3 years and AI will be better than humans at absolutely everything... the investors have been saying this for the last 3 years. Like I remember even in our seed round... a lot of the top firms in Silicon Valley just thought that the market would dissipate within a couple of years." 00:41:47 He treats this as a gift: "Being contrarian reduces the amount of competition in the market. I'm glad our market is not nearly as competitive as the foundation model business." 00:41:47

API revenue is lower quality than investors think

Foody calls API businesses the most vulnerable layer. Brendan Foody: "The switching costs for APIs are so low that I think investors overrate some of the revenue quality of API businesses... we switch between APIs on a daily cadence based on whatever models performing the best on our evals for how we run the business. I think every company will do that as they mature and have scaled inference bills." 00:40:37

Software layers and GPU margins get commoditized, even at celebrated companies

His claim is that large parts of every AI business get commoditized. Brendan Foody: "If everyone's buying GPUs in the inference layer, then the GPUs get commoditized. If the app layer companies are building software layers, those will get commoditized as it becomes cheaper and cheaper to build software." 00:37:05 On GPU margins: "As much as people try to tell this nice story of how there's great margins, like of course when you buy a bunch of GPUs and those appreciate like 50%, you'll have like good margins for a period of time." 00:38:46

Don't watch your competitors, because it's a recipe for second place

Where many founders obsess over rivals, Foody calls it a trap. Brendan Foody: "Many companies fall too much into the trap of looking at what their competitors do, which becomes mimetic. Like it's a really good way to be second place in a market, right? And we have no interest in being second place in a market." 00:18:42

Tax policy: taxing jobs for the bottom half of earners is irrational

Foody's college research thesis argued the largest positive externality is job creation, yet governments tax jobs. Brendan Foody: "The bottom half of Americans constitute 3% of federal government tax revenue and yet it's a giant burden on them... So the thesis was just eliminate taxes for the bottom half of Americans because that will incentivize more job creation." 00:25:16

3. Companies Identified

Mercor

AI training data and RL environment company; talent platform, Studio data platform, AI project manager. Mentioned as the subject. Brendan Foody: "We've grown by more than 5x on a mid-9 figure base." 00:00:47 On expert pay: "pay out over $4 million a day" and "to date we've paid out just about a billion dollars historically." 00:02:42 On interviews: "we've done over 7 million interviews... we have now created the fastest growing job category in history." 00:12:28 Also notable: "we're spending almost like a million dollars a day on inference." 00:46:57

Deeptune

New York-based startup building clones of enterprise apps (Salesforce, Adobe, Google Workspace) for RL environments; acquired by Mercor. Mentioned as the fix for Mercor's largest bottleneck. Brendan Foody: "Our largest bottleneck was how do we get the apps, the clones of Salesforce and Adobe and Google Workspace, that we can have all of the experts populate with the relevant content." 00:28:47 "One of the customers that we work with actually suggested proactively that Deeptune was their top apps partner." 00:28:47 It had "just raised his Series A from a16z." 00:30:27

Seepel

A small startup founded by Robbie and Cat, acquired by Mercor as its first acquisition. Brendan Foody: "We also acquired a company called Seepel, founded by Robbie and Cat, who are also spectacular." 00:30:27

Harvey

AI platform for legal work. Mentioned as a model for app-layer moats and as a partner. Brendan Foody: "Harvey has a world-class brand, certainly top in the law category and maybe more generally across application layer startups." 00:37:05 On the partnership: "The work that we do with them corresponds to starting with great experts across all the granular domains that they care about." 00:35:20 Harvey was also cited as past the $100M inference threshold where post-training economics check out. 00:44:45

Cognition

AI coding agent company, led by Scott Wu. Mentioned as an early Mercor customer that became large. Brendan Foody: "Some of them became large companies like Cognition." 00:20:12 Molly O'Shea noted that Cognition is staying independent while "most of the companies in coding agents have gotten acquired." 00:33:28

Fireworks (and Together)

Inference providers. Foody named Fireworks as executing correctly on custom models. Brendan Foody: "The primary one that I think the inference companies need to get right and Fireworks is executing really well on, is building custom models." 00:38:46

Cursor

AI coding product. Used as an example of a company past $100M inference spend. Brendan Foody: "Once they're past $100 million in inference, like a Harvey or a Cursor... sometimes, you know, hundreds or billions in the case of Cursor, in training your own models." 00:46:57

Anthropic and OpenAI

Frontier labs. Foody's view of their potential: "I think both of them could very well be $10 trillion plus companies." 00:44:46 Their moats are economies of scale in compute and data plus brand, with "some flywheels associated with a given model helping them to improve their subsequent model more." 00:38:46

Scale AI

Data-labeling competitor; Alex Wang and core team went to Meta. Brendan Foody: "Alex and the core team went over and then a lot of the rest of the team stayed at Scale and Scale pivoted most of their business to just focus on Meta. I don't think we would ever do something like that." 00:32:25

Stripe (Stripe Connect)

Payments infrastructure for Mercor's expert payouts. Brendan Foody: "It uses Stripe Connect as the back end." 00:04:10

Brex

Sponsor; finance platform used by Sourcery and cited as used by Vercel, OpenAI, Anthropic, Granola, and Deepgram. 00:10:54

Goldman Sachs

Used as an example of why enterprises won't let labs train on their data. Brendan Foody: "The reason that a company like Goldman Sachs exists is that they have a set of competitive advantages associated with the people, the processes, the documents, etc. in their business." 00:44:46

Amazon

Cited for Bezos's long-term orientation. Brendan Foody: "Being so principled about what are the things that matter at the very beginning of Amazon and how those map onto the product strategy decades later, is a lot more challenging than it looks." 00:23:31

Other companies mentioned in passing

Meta: acquirer of Scale AI's core team; "Meta still uses other vendors." 00:32:46
Adobe: referenced in Dylan Field's scrapped acquisition. 00:18:15
Figma: referenced for personal projects used in expert evaluation and Dylan Field's story. 00:12:28
a16z: led Deeptune's Series A. 00:30:27
Upfront Summit: where Foody and O'Shea crossed paths. 00:21:34
Zone, VCX by Fundrise, Public, Deel, Vercel, Granola, Deepgram: sponsors/customers named in ad reads. 00:27:15

4. People Identified

CEO and co-founder of Mercor, college dropout, early 20s. The guest. Notable for building "the fastest growing job category in history" and for a 100-hour-week operating cadence. "I'm in the office until 10:00 p.m. and I rinse and repeat 6 days a week, and then Sundays I work from home." 00:51:32

Adarsh (Adarsh Hiremath)

Co-founder of Mercor. Mentioned as Foody's co-founder who also turned down cash-out opportunities. Brendan Foody: "Adarsh and I have had opportunities where we could walk away with a lot of cash and never work again... but that's not what motivates us." 00:00:00 Also part of the original trio that built the first AI interviewer. 00:12:28

Shrey

Co-founder of Mercor, part of the team that built the first AI interviewer in April 2023. Brendan Foody: "We built the first AI interviewer when Adarsh, Shrey, and I were in college in April of 2023." 00:12:28

Elon Musk

Mentioned as Foody's greatest entrepreneur by results. Brendan Foody: "Elon is the greatest entrepreneur by results... the mission orientation of really enduring so much pain, because he cares about building these companies that shape the trajectory of a really important market." 00:23:23 Foody also cites his "chewing glass and staring into the abyss" line. 00:20:12

Jeff Bezos

Foody's hero; embodies long-term thinking. Brendan Foody: "He really embodies this long-term thinking more so than almost anyone... the ability of him to create an entrepreneurial culture." 00:23:31 Bezos reposted Foody's tax-policy tweet. 00:25:16

Sam Altman and Dario Amodei

Foody's founder heroes. Brendan Foody: "Sam and Dario are both spectacular. I often feel bad that both of them will always get so much hate... you're on top of the world and then like you're the America's worst enemy, and then all of a sudden you're back." 00:23:31

Alex Wang

Founder of Scale AI, now at Meta. Brendan Foody: "He was one of the most generational entrepreneurs of the last decade and so it's been great to see what he's been able to accomplish at Meta." 00:32:58

Scott Wu

CEO of Cognition. Brendan Foody: "Scott is so fun to listen to." 00:33:26 Molly O'Shea noted he wants to build a massive independent company.

Tim (Deeptune founder)

Founder of Deeptune. Brendan Foody: "I got to know Tim, the founder, better, and he was spectacular. And that's obviously half of any of these things, right? Just like who are the people that you want to work with for the next decade." 00:28:47

Robbie and Cat (Seepel founders)

Founders of Seepel, acquired by Mercor. Brendan Foody: "Founded by Robbie and Cat, who are also spectacular." 00:30:27

Winston Weinberg, Gabe Pereyra, Niko Grupen (Harvey)

Harvey's leaders, all previously guests on Sourcery. Foody: "I love those guys, too." 00:34:40 Winston supplied an AI data sovereignty question. 00:44:46

Alfred Lin

Sequoia partner whose "shock absorber" philosophy Foody endorsed. Brendan Foody: "What Alfred was saying is super accurate, that when everyone feels like they're on top of the world, that's when you need to be cracking down on what are all the things going wrong. And when things are not going well... that's when you want to be the cheerleader." 00:22:00

Dylan Field

Figma CEO; his post-failed-Adobe-deal "chip on shoulder" story framed Molly's "clean fuel" question. Molly O'Shea: "He was just running on pure clean fuel... Then that happened, got a major chip on his shoulder and he was like, 'No, I'm going to fucking win.'" 00:18:16

Jensen Huang, Alex Karp, Satya Nadella

Cited as leaders talking about AI sovereignty. Brendan Foody: "There's a reason that Jensen and Alex Karp and Satya are all talking about this." 00:44:46

5. Operating Insights

Over-invest in the unglamorous function that failed a competitor

Mercor's world-class payments team was born from watching a partner botch payouts. Brendan Foody: "We were working with another player in the space that was having a lot of payout issues very publicly, where we were providing talent to them and we got flooded with support tickets that we hired over 1,000 people for them... they were not getting paid and struggling to put food on the table for their families." 00:04:19 The conclusion was to build a direct lab relationship and "over invest in how do we have a world-class payments team that never makes mistakes." Takeaway: a competitor's operational failure is a roadmap for your own reliability moat.

Scale evaluation with AI to fix talent-density differences

The first AI interviewer (7-8 seconds latency in April 2023) turned hiring into a scalable data-collection problem. Brendan Foody: "Hiring Stanford AI PhDs, you can probably meet with five people and one of them is going to be pretty good, versus... remote place in India, you need to meet with sometimes hundreds or 1,000 people to find like the really world-class engineers, but they are there." 00:12:28 Pulling referrer, GitHub, and Figma into context lets you find top talent in lower-density pools.

Strip vanity metrics; measure the inputs that compound

Foody removed a weekly revenue chart from circulation and redirected attention to customer happiness, expert happiness, and expert utilization. Brendan Foody: "The company should be looking at... all of these different things that like actually drive creating value in the world and how that compounds." 00:22:00

Alternate between cheerleader and critic depending on the cycle

Foody adopts Alfred Lin's rule: be the stern critic when things are going great and the cheerleader when they're rough. 00:22:00 This regulates company emotional state against recency bias.

Use evals as the daily routing layer for model vendors

Mercor treats model vendors as swappable. Brendan Foody: "We switch between APIs on a daily cadence based on whatever model's performing the best on our evals." 00:40:37 Build a private eval first, then route inference on it; this also applies to how Foody advises enterprises to "start with the eval." 00:48:25

6. Overlooked Insights

The "tax on jobs" thesis hints at a worldview about where Mercor's economics come from

In a seemingly off-topic aside, Foody argues "the largest positive externality in the economy is creating jobs." 00:25:16 Combined with the claim that Mercor created "the fastest growing job category in history" 00:12:28 and is hiring aggressively while others say AI displaces jobs, it suggests Foody frames Mercor less as a data vendor and more as a labor-market layer: expert-teaching becomes the dominant form of knowledge work. His closing line reinforces it: "Everyone says, 'Oh, like we're going to displace all the jobs,' but all of the AI companies, us included, are hiring as quickly as possible." 00:53:43 The implicit bet is that the human share of economic output is re-routed, not eliminated, through a new job category Mercor intermediates.

A customer introduced the acquisition target, and the reason is a quiet signal about lab behavior

Foody mentioned in passing that a Mercor customer (a lab) "proactively suggested that Deeptune was their top apps partner and they should just give us apps that we could populate." 00:28:47 This reveals that labs are actively orchestrating a supply chain of RL-environment vendors (apps from one, worlds and tasks from another) and nudging vendors to consolidate. That implies labs value integrated, full-stack environment suppliers and will shape which data vendors survive, which matters for anyone evaluating the data and RL-environment category or its component suppliers.