Is Olam Labs profitable?

No revenue or profitability data exists for Olam Labs as of July 2026. The company is a 2-person, pre-seed YC S26 startup with no disclosed customers, and its products are still labeled 'Alpha' with no ranked leaderboard released yet.

Has Olam Labs had layoffs?

No layoffs are reported. Olam Labs is a newly founded 2026 startup with a 2-person team based in San Francisco, part of Y Combinator's Summer 2026 batch, and has no hiring history to reduce as of July 2026.

Could AI replace Olam Labs' own business?

Olam Labs' core method, game-based LLM evaluation, is already open academic art via clembench (since 2023), GameEval, GVGAI-LLM, and V-MAGE, meaning frontier labs and open frameworks could replicate its orchestration approach without needing the startup itself.

Who is Olam Labs' biggest competitor?

Mercor, valued at $10B in October 2025, acquired simulation-eval startup Deeptune in July 2026, directly entering Olam Labs' niche. Scale AI (49% Meta-owned) and Surge AI (reportedly $1B+ revenue) also operate at incumbent scale in adjacent data-eval markets.

Olam Labs is privately held with no independent parent company; it is backed by Y Combinator as part of the Summer 2026 (S26) batch, based in San Francisco. No funding round or valuation beyond YC's standard batch investment has been publicly disclosed as of July 2026.

Does Olam Labs have a working product yet?

Yes, called Social Arena: Catan-style, Risk-style, poker-style, and Codenames-style games where humans unknowingly play against AI agents. As of July 2026 all games are labeled 'Alpha,' still collecting data before any leaderboard or benchmark release.

YC Summer 2026 · United States

Olam Labs

Two-person startup testing AI models by having humans unknowingly play games against them.

Protection under test

Assessed 27 July 2026 · Evidence as of 27 July 2026

Something worth protecting exists and is being tested right now.

On the recordSince 27 July 2026 · Checkable from Q3 2026 · Every call on the record Checkpoint: YC S26 Demo Day and seed fundraising

The call

Olam Labs has an idea that academic frameworks already do for free and that funded rivals are now buying up, with no ranked product yet to show it collects data those rivals cannot.

Olam Labs uses games to test how well AI systems work, which is a smarter way to evaluate AI than standard tests. The problem is that universities, other startups like Scale AI and Surge AI, and even OpenAI have all launched similar game-based testing tools since 2023, so the basic idea is no longer unique. Olam's only real advantage is data from real people playing against AI, but that data collection is still in early stages with no public results to show. The one thing that could change everything is if they launch a public leaderboard that becomes the standard place where AI companies want their models ranked, which would give them real stickiness before bigger competitors fully copy them.

Why

AI pressure: AI2's olmo-eval and frontier labs' internal evaluation tools. Labs like AI2 now ship their own eval workbenches in house, cutting the need to pay an outside game-based tester. Could AI do the core job? Yes, likely.

Bigger players: Mercor (after acquiring Deeptune), Scale AI, Surge AI. Mercor bought a simulation-eval startup in July 2026 and Scale AI already sells eval services at scale, boxing in a two-person shop. Could a giant copy it? Yes, easily.

What protects it. If Olam can gather human-vs-AI social behavior data unknowingly, that no academic benchmark captures, before rivals copy the game harness.

What would change this call

Whether OpenAI, Anthropic, or Meta names Olam Labs in a published eval methodology within a year. Would show a frontier lab values Olam's data enough to cite or pay for it.
Olam Labs publicly launches a ranked leaderboard or benchmark from Social Arena. Would show the product has moved past alpha and is gathering usable data at scale.
Mercor, Scale AI, or Surge AI announce a game based behavioral eval feature within a year. Would show funded incumbents are closing the gap Olam relies on for defensibility.

Key numbers

Team size: 2 people, San Francisco (2026-07)
Mercor valued at $10B after $350M raise (2025-10)
AI2 shipped olmo-eval workbench (2026-06)

Named risks

Core game-based evaluation method is open academic art, not proprietary, easily replicated. Severity: high
Funded incumbents (Mercor, Scale, OpenAI, AI2) are actively acquiring or building the same niche. Severity: high
Product remains unranked 'Alpha' with no leaderboard, benchmark, or disclosed customers as of July 2026. Severity: high
No disclosed funding round beyond standard YC investment, limiting runway to compete. Severity: medium
2-person team faces execution risk against 15+ employee peers like HUD. Severity: medium

Rivals, as read in July 2026

Rival	Direction	Evidence
Mercor	losing to	Acquired sim-eval startup Deeptune, July 2026, directly entering Olam's niche
Scale AI	losing to	49% Meta-owned, operates human-panel data at industrial scale vs Olam's 2-person team
OpenAI (Promptfoo)	losing to	OpenAI acquired eval tool Promptfoo March 2026, absorbing eval-tooling niche
AI2 (olmo-eval)	contesting	Shipped open eval workbench olmo-eval June 2026, free alternative to proprietary eval startups
clembench (academic)	contesting	Open dialogue-game LLM eval framework active since 2023, non-proprietary precedent
Surge AI	losing to	Reportedly $1B+ revenue building realistic RL environments for frontier model evaluation

Questions people ask

Is Olam Labs profitable?: No revenue or profitability data exists for Olam Labs as of July 2026. The company is a 2-person, pre-seed YC S26 startup with no disclosed customers, and its products are still labeled 'Alpha' with no ranked leaderboard released yet.
Has Olam Labs had layoffs?: No layoffs are reported. Olam Labs is a newly founded 2026 startup with a 2-person team based in San Francisco, part of Y Combinator's Summer 2026 batch, and has no hiring history to reduce as of July 2026.
Could AI replace Olam Labs' own business?: Olam Labs' core method, game-based LLM evaluation, is already open academic art via clembench (since 2023), GameEval, GVGAI-LLM, and V-MAGE, meaning frontier labs and open frameworks could replicate its orchestration approach without needing the startup itself.
Who is Olam Labs' biggest competitor?: Mercor, valued at $10B in October 2025, acquired simulation-eval startup Deeptune in July 2026, directly entering Olam Labs' niche. Scale AI (49% Meta-owned) and Surge AI (reportedly $1B+ revenue) also operate at incumbent scale in adjacent data-eval markets.
Who owns Olam Labs?: Olam Labs is privately held with no independent parent company; it is backed by Y Combinator as part of the Summer 2026 (S26) batch, based in San Francisco. No funding round or valuation beyond YC's standard batch investment has been publicly disclosed as of July 2026.
Does Olam Labs have a working product yet?: Yes, called Social Arena: Catan-style, Risk-style, poker-style, and Codenames-style games where humans unknowingly play against AI agents. As of July 2026 all games are labeled 'Alpha,' still collecting data before any leaderboard or benchmark release.

Timeline

2026-07-22YC directory lists Olam Labs at 2 employees under Reinforcement Learning category
2026-07Mercor acquires simulation-eval startup Deeptune, entering Olam Labs' niche
2026-07Olam Labs confirmed as part of Y Combinator Summer 2026 (S26) batch, San Francisco
2026-06-16YC P26 batch breakdown: 137 of 196 companies building LLM agents
2026-06-12AI2 ships olmo-eval, an open evaluation workbench for LLM development
2026-03OpenAI acquires Promptfoo, consolidating AI eval tooling market
2025-10Mercor reaches $10B valuation after $350M Series C round
2025-06-19Meta takes ~49% stake in Scale AI via $14.3B investment

Revision history

27 July 2026

First assessed. Protection under test. The text above is fixed as of this date and is not edited afterwards.

In the same band

Prized Truffle Daqstra