Is Hillclimb profitable in 2026?

No profitability data is public. Hillclimb is a 4-person YC Fall 2025 startup with an unverified funding estimate of $500K-$1M and no confirmed revenue figures as of mid-2026, per its YC company page dated 2026-06-18.

Could AI replace Hillclimb's business model?

Possibly. Hillclimb sells human-expert math training data, but math is the domain most automatable via synthetic, verifier-checked self-play data, and frontier labs are increasingly building RL environments in-house, per Epoch AI reporting dated 2026-01-12.

Who is Hillclimb's biggest competitor?

Mercor, valued around $10 billion after its October 2025 Series C with over $350 million raised and roughly 30,000 experts on its marketplace, dwarfs Hillclimb's 4-person team and one named customer, Nous Research.

Hillclimb is privately held, founded in 2025 by Jun Park, Ibrakhim Ustelbay, and Pruthvi Rajaghatta, and went through Y Combinator's Fall 2025 batch. It claims backing from Garry Tan and angel investors from OpenAI, Anthropic, and DeepMind, per a March 2026 job posting.

How much funding has Hillclimb raised?

Unconfirmed. Aggregator estimates suggest roughly $500K-$1M, consistent with a standard YC pre-seed check, but no priced round or valuation has been publicly reported as of July 2026, and figures conflict across sources.

What is Hillclimb's flagship product or deal?

Nomos 1, a 30B-parameter math-specialized model built on Qwen3-30B with Nous Research, released December 9, 2025, and explicitly credited to Hillclimb's founders and named math experts on Hugging Face.

YC Fall 2025 · United States

hillclimb

Supplies elite human math experts and RL training data to AI labs building reasoning models.

Losing its protection

Assessed 27 July 2026 · Evidence as of 27 July 2026

The structural protection is eroding.

On the recordSince 27 July 2026 · Checkable from H2 2026 · Every call on the record Checkpoint: Expected seed or Series A raise following YC F25, which would test investor confidence in the vendor model

The call

A four-person team with one named customer sits between giants buying up rivals and labs learning to generate their own math training data.

Hillclimb hires elite mathematicians to create training data that teaches AI systems to solve hard math problems. The problem is that larger companies like Mercor and Surge can simply hire the same math experts themselves, and AI labs are starting to build their own math training inside their own walls instead of buying it from outside vendors. The one thing that could save Hillclimb is if it builds a unique library of math problems and solutions that no one else has, or if it grows fast enough to become one of the last few winners before the market consolidates.

Why

AI pressure: Frontier labs' self-play + Lean-verified synthetic math data. Math is the easiest domain to auto-verify, so labs can generate their own checked training data and skip human experts entirely. Could AI do the core job? Yes, likely.

Bigger players: Mercor. Mercor, valued near $10B, is buying up specialized data vendors like Sepal AI and Deeptune, showing it can absorb niches like elite math easily. Could a giant copy it? Yes, easily.

What protects it. Its named partnership with Nous Research on Nomos 1, with public co-authorship credit for contributing mathematicians, gives it a visible track record labs can point to.

What would change this call

Hillclimb announces a named frontier-lab customer beyond Nous Research within 12 months. Would show it can win business from major labs, not just small open labs.
Mercor or Surge AI acquires or launches a math-expert RL data vendor or team. Would show incumbents replicating Hillclimb's elite-math niche via acquisition.
A frontier lab publishes a Lean-verified or self-play synthetic math training pipeline replacing human annotators. Would validate that Hillclimb's human-expert math wedge is becoming automatable.

Key numbers

4 employees total (2026-06-18)
Mercor valued at ~$10B, $350M+ raised (2025-10)
Nomos 1 math model shipped with Nous Research (2025-12-09)

Named risks

Frontier labs vertically integrating RL environments in-house, cutting vendor demand. Severity: high
Math is the most verifier-friendly domain for synthetic self-play data to replace human experts. Severity: high
Traction rests on a single named customer, Nous Research, a small open lab. Severity: high
Sector consolidation expected to narrow ~20-38 vendors to 3-5 winners by 2030. Severity: medium
4-person team faces well-funded incumbents actively acquiring specialist niches. Severity: medium

Rivals, as read in July 2026

Rival	Direction	Evidence
Mercor	losing to	Mercor at ~$10B valuation, ~30,000 experts, acquired Sepal AI Feb 2026 and Deeptune July 2026
Surge AI	losing to	Surge AI in talks for ~$25B valuation with $1B+ revenue as of June 2026
Handshake AI	losing to	Handshake AI reached ~$280M annualized revenue by Aug 2025
Scale AI (Meta)	losing to	Scale AI absorbed into Meta via ~$14.3B deal in 2025, part of Big Four controlling 75%+ share
Nous Research	contesting	Sole named partnership; co-released Nomos 1 math model Dec 2025, but Nous is a small open lab not a frontier major

Questions people ask

Is Hillclimb profitable in 2026?: No profitability data is public. Hillclimb is a 4-person YC Fall 2025 startup with an unverified funding estimate of $500K-$1M and no confirmed revenue figures as of mid-2026, per its YC company page dated 2026-06-18.
Could AI replace Hillclimb's business model?: Possibly. Hillclimb sells human-expert math training data, but math is the domain most automatable via synthetic, verifier-checked self-play data, and frontier labs are increasingly building RL environments in-house, per Epoch AI reporting dated 2026-01-12.
Who is Hillclimb's biggest competitor?: Mercor, valued around $10 billion after its October 2025 Series C with over $350 million raised and roughly 30,000 experts on its marketplace, dwarfs Hillclimb's 4-person team and one named customer, Nous Research.
Who owns Hillclimb?: Hillclimb is privately held, founded in 2025 by Jun Park, Ibrakhim Ustelbay, and Pruthvi Rajaghatta, and went through Y Combinator's Fall 2025 batch. It claims backing from Garry Tan and angel investors from OpenAI, Anthropic, and DeepMind, per a March 2026 job posting.
How much funding has Hillclimb raised?: Unconfirmed. Aggregator estimates suggest roughly $500K-$1M, consistent with a standard YC pre-seed check, but no priced round or valuation has been publicly reported as of July 2026, and figures conflict across sources.
What is Hillclimb's flagship product or deal?: Nomos 1, a 30B-parameter math-specialized model built on Qwen3-30B with Nous Research, released December 9, 2025, and explicitly credited to Hillclimb's founders and named math experts on Hugging Face.

Timeline

2026-07-24Report pegs RL-environment sector at $8.5B revenue, Big Four holding over 75% share
2026-07Mercor acquires Deeptune, expanding specialist vendor rollup
2026-06-18YC company page lists Hillclimb team size at 4, status active
2026-03-19Job posting claims backing from Garry Tan and top-lab angel investors
2026-02Mercor acquires Sepal AI, targeting specialized data capability
2026-01-26Analysts predict RL-environment vendors will narrow from ~30 to 3-5 winners by 2030
2025-12-09Nous Research releases Nomos 1, math model co-trained with Hillclimb
2025-11-14Hillclimb publicly launches on YC's launch platform
2025-09Anthropic reportedly discusses spending over $1B/year on RL environments
2025-04Nous Research releases Atropos, open async RL-environment infrastructure

Revision history

27 July 2026

First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.

In the same band

Nexus Terrain Metorial