On this page
Snapshot
What it is
- Detail
- Evals for robots: open-source tools and independent benchmarks that measure how well robots do real-world jobs[1]
- As of
- Nov 2026
Founded
- Detail
- 2026[1]
- As of
- Nov 2026
Headquarters
- Detail
- San Francisco, California[1]
- As of
- Nov 2026
Leadership
- Detail
- Jay Chooi, founder and CEO[1]
- As of
- Nov 2026
Employees
- Detail
- 3[1]
- As of
- Nov 2026
Legal entity
- Detail
- A public benefit corporation, "with a legal duty to independently evaluate the robotics capabilities of frontier AI systems"[2]Self-reported
- As of
- Sep 2026
Accelerator
- Detail
- Y Combinator, Summer 2026 batch[1]
- As of
- Nov 2026
Leadership
CEO
- Person and background
- Jay Chooi, founder and CEO[1]
- As of
- Nov 2026
Funding
Seed
- Detail
- $10M seed (Sep 2026) led by Initialized Capital, with Notable Capital, Decasonic, Y Combinator, Halcyon Futures "and many others"[2]
- As of
- Sep 2026
Products and models
Business model
- Detail
- Evaluates how well frontier AI models control real robots, as an independent third-party auditor; says frontier labs do not direct its evaluations or control its methodology or public results[2]Self-reported
- As of
- Sep 2026
Software
- Detail
- Inspect Robots, an open-source evaluation harness (GitHub)[3]
- As of
- Nov 2026
Published evaluations
- Detail
- StationeryBench (Sep 10, 2026): five two-armed manipulation tasks with desk stationery, run on the same YAM arms. OpenAI's GPT-6 Astra, driven by Robocurve's Inspect Robots agent policy, completed 7 of 100 trials and Ai2's MolmoAct2 none. Robocurve's own evaluation, on its openai.robocurve.org subdomain; the page claims no partnership with OpenAI[4]Self-reported
- As of
- Sep 2026