Mirendil
Sell the self-improving AI the big labs keep in-house, as an AI researcher platform that designs experiments, tunes settings, evaluates models and runs the next training round, so outside organizations can build their own models ('AI for AI for science').
- Founded
- December 2025
- Headquarters
- San Francisco, CA
- Founders
- Behnam Neyshabur, Harsh Mehta
- Pushes against
- Frontier labs that use AI-automated research internally and forbid outsiders from doing so in their terms of service.
- Now
- Launched with the seed on 2026-06-24, six months after both founders left Anthropic (after Claude Opus 4.5). No product released yet as of 2026-10-04.
Key people
- Behnam Neyshabur (CEO; ex-Anthropic, ex-Google Gemini reasoning)
- Harsh Mehta (ex-Anthropic, ex-Google)
Funding
- 24 June 2026
- Amount
- $200M seed
- Valuation
- $1B
- Lead
- Co-led by Andreessen Horowitz and Kleiner Perkins, with Nvidia
Every funding round in the atlas
Sources
- thenextweb.com/news/mirendil-200m-seed-ai-research
- cnbc.com/2026/04/28/meta-google-big-tech-staff-ai-labs-investors.html
- a16z.com/announcement/investing-in-mirendil/
- siliconangle.com/2026/06/25/mirendil-raises-200m-speed-scientific-research-ai/
- letsdatascience.com/news/mirendil-raises-200m-to-build-self-improving-ai-platform-c564b966
This profile was partly confirmed by its sources on 6 October 2026. Some of its claims could not be checked. How we check