Sourcery with Molly O'Shea · the Podbrain notes ·
3 min read

Mercor Raises $750M at $20B Valuation | CEO Brendan Foody

In this episode of Sorcery, host Molly interviews Brendan, co-founder and CEO of Mercor, a leading AI talent and reinforcement learning (RL) data platform. Brendan outlines Mercor's hypergrowth trajectory since dropping out of college, scaling from an initial revenue base of under $900,000 to distributing over $5...

Sourcery with Molly O'Shea Sourcery with Molly O'Shea
Subscribe to Notes Upgrade
Sourcery with Molly O'Shea episode thumbnail: Mercor Raises $750M at $20B Valuation | CEO Brendan Foody
Sourcery with Molly O'Shea
Key Takeaways
  1. 01

    Mercor's talent payouts expanded from $1 million annually two years ago to over $5 million daily, totaling nearly $1 billion paid out to date.

  2. 02

    Over 70% of Mercor's 5 million experts are sourced through direct referrals, while the platform has conducted over 7 million AI-powered candidate interviews.

  3. 03

    "We would rather work 100 hours a week to solve a super important problem for the world than take high payout exits." - Brendan

  4. 04

    Brendan rejects claims that human-generated AI data is unsustainable, citing Jevons paradox as expert productivity increases by 3-5x using AI tools.

  5. 05

    Enterprise AI spending drives post-training model development once annual inference costs exceed $100 million, making proprietary data and benchmark ownership critical.

  6. 06

    Mercor expanded its RL environment capabilities by acquiring Deep Tune to integrate enterprise application layers like Salesforce, Adobe, and Google Workspace.

Get the latest ideas from Sourcery with Molly O'Shea.

Plus the best new takeaways about artificial intelligence from other top podcasts — read in minutes, not hours.

or

By continuing, you agree to Podbrain's Terms and Privacy Policy.

These notes may contain occasional inaccuracies. Learn how Podbrain notes are made

In this episode of Sorcery, host Molly interviews Brendan, co-founder and CEO of Mercor, a leading AI talent and reinforcement learning (RL) data platform. Brendan outlines Mercor's hypergrowth trajectory since dropping out of college, scaling from an initial revenue base of under $900,000 to distributing over $5 million per day in payouts to more than 5 million domain experts globally.

The discussion explores Mercor's three-part platform architecture—comprising talent matching, automated Studio data environments, and autonomous AI project management agents. Brendan explains how Jevons paradox increases demand for human expert feedback as models advance, the strategic necessity of acquiring application layer tools like Deep Tune, and how Fortune 500 enterprises navigate scaling inference costs from $10 million to over $100 million. Additionally, Brendan shares his perspectives on organizational culture, execution velocity, founder routines, and economic tax policies on labor.

Scaling Mercor's RL Data and Talent Operations

Mercor has achieved over 5x growth since September of the previous year on a base exceeding $900,000, paying out over $5 million daily to experts.

The company has disbursed nearly $1 billion in total payouts to date using Stripe Connect as its backend payments infrastructure.

Mercor maintains a network of over 5 million active job seekers, with 70% of experts onboarded through direct recommendations.

Brendan and co-founder Adarsh Suri created their first AI voice interviewer in April 2023, reaching a milestone of over 7 million candidate interviews completed.

"We'd rather work 100 hours a week solving a super important problem for the world than get rich and stop working." - Brendan

Architecture of Reinforcement Learning Environments

Mercor's RL environments consist of three distinct components: enterprise applications, content-populated worlds, and structured evaluation tasks.

AI assistant tools allow human experts to become 3x to 5x more productive per hour when building specialized dataset environments.

Brendan invokes Jevons paradox to explain why expert feedback demand grows exponentially rather than shrinking as AI models become more capable.

"99% of knowledge still exists inside human minds, and models require human feedback at the frontier to measure where they fail." - Brendan

Strategic M&A and Enterprise AI Inference Scaling

Mercor acquired Andreessen Horowitz Series A-funded Deep Tune to integrate enterprise application interfaces like Salesforce, Adobe, and Google Workspace.

The transaction marks Mercor's second major acquisition following its deal for talent startup Seople, helping establish its New York presence.

Companies spending $10 million to $100 million on AI inference typically allocate roughly 10% of their budget toward model evaluation datasets.

Enterprise inference spending exceeding $100 million annually makes custom post-training model development economically viable.

"Sovereign AI data will be critical because Fortune 500 companies will not allow frontier model labs to train on their proprietary internal data." - Brendan

Founder Execution, Company Culture, and Work Routines

Mercor operates on three core principles: uncompromising hiring standards, long-term strategic forward vision, and customer obsession.

"If you focus on watching competitors, you become a copycat, which is a great way to end up second in the market." - Brendan

Brendan works six days a week in the office from 9:00 AM to 10:00 PM, splitting time equally between recruiting, client relations, and product R&D.

Brendan advocates eliminating federal income taxes on the bottom 50% of earners, noting they account for only 3% of revenue while discouraging job creation.

Sourcery with Molly O'Shea
From Sourcery with Molly O'Shea. Get a note like this from every new episode.
Subscribe to Notes Upgrade

These notes may contain occasional inaccuracies. Learn how Podbrain notes are made

0 / 0
Link copied