Mirendil Secures $200M Seed to Automate AI Research
Mirendil, a San Francisco AI startup founded by former Anthropic researchers Behnam Neyshabur and Harsh Mehta, raised 1 billion valuation (announced June 26, 2026).
Mission
Mirendil aims to automate the AI research process itself — enabling autonomous experimental design and iterative engineering so institutions outside major labs can conduct frontier R&D without massive internal engineering teams.
Founders (previously at Alphabet and Google DeepMind) argue AI research concentration in a few tech firms bottlenecks global innovation. The platform targets compressing multi-month research cycles into days through automated iteration.
Funding and Backers
Co-led by Andreessen Horowitz and Kleiner Perkins, with NVIDIA participation.
Team and Differentiation
Team includes talent from xAI, OpenAI, and MIT. Unlike Safe Superintelligence or Thinking Machines Lab (focused on model development/superintelligence), Mirendil positions as infrastructure for research automation — a platform layer for external organizations such as university biology labs and enterprise ML teams.
AI infrastructure revenue is projected to reach $1.2 trillion by 2030.
Market Context
Part of a wave of high-profile ventures from alumni of established AI labs. Success depends on proving automated research can replicate institutional knowledge and proprietary infrastructure held by larger competitors.