- Mirendil, a San Francisco-based AI startup founded by two former Anthropic researchers, has raised $200 million in seed funding at a $1 billion valuation.
- The founders left Anthropic in December 2025 to build AI systems that automate AI research itself, aiming to bring frontier R&D within reach of universities and research institutions beyond the largest labs.
- Andreessen Horowitz and Kleiner Perkins led the round, with NVIDIA among the participating investors.
Behnam Neyshabur and Harsh Mehta left Anthropic in December 2025, months after joining the company in late 2024, and have now raised $200 million at a $1 billion valuation for their new startup, Mirendil. The raise is one of the largest seed rounds in AI history.
The San Francisco-based company wants to build AI that can do what AI researchers do: design experiments, iterate on engineering problems, and improve AI systems over time with progressively less human input.
The round was co-led by Andreessen Horowitz and Kleiner Perkins, with participation from NVIDIA. It is the pair’s first company together, and Mirendil has yet to ship a product.
The global AI infrastructure market ended 2025 with $337 billion in revenue and is projected to reach $1.2 trillion by 2030, growing at a 28% compound annual growth rate. AI captured close to 50% of all global venture funding in 2025, with $202 billion invested across the sector, per Crunchbase, up more than 75% year-on-year.
Demand is being driven by the gap between the research capabilities of the largest AI labs and everyone else: universities, hospitals, and enterprises that want to build specialised AI but lack the teams and infrastructure to do it
The founders and what they are building
Neyshabur, who spent 5.5 years at Alphabet co-leading Gemini’s reasoning research before joining Anthropic in December 2024, first met Mehta in 2019, when both were at Google. The two spent the following years working on large-scale model training and evaluation before leaving Anthropic together at the end of 2025. Their argument is simple: the ability to run frontier AI research should not be confined to a handful of technology companies.
Mirendil, founded in early 2026, works by training specialised models on frontier AI research tasks, including experimental design, hyperparameter search, model evaluation, and iterative training, and packaging those capabilities as a platform that external organisations can deploy on their own problems. A university biology lab, for example, could use it to build and refine a model for drug-target identification without a dedicated machine-learning engineering team. The ambition is to compress research cycles that currently take months of human work into days of automated iteration.
The founding team also includes Shayan Salehian, an early member of xAI, and Tara Rezaei, an MIT graduate, as well as about 20 researchers and engineers recruited from Anthropic, xAI, Google DeepMind, and OpenAI.
How Mirendil differs from the wave of frontier lab spinouts
Tech Funding News has tracked a wave of heavily funded AI labs founded by alumni of the same companies that Mirendil’s team came from, but none are targeting the exact same layer.
Safe Superintelligence, led by former OpenAI chief scientist Ilya Sutskever, has raised $6 billion at a $32 billion valuation with a singular focus on safe superintelligence and no commercial product. Thinking Machines Lab, founded by former OpenAI chief technology officer Mira Murati, raised $2 billion at a $12 billion valuation before a reported $50 billion fundraise collapsed.
Closer to Mirendil’s territory, Recursive Superintelligence raised $650 million at a $4.65 billion valuation to build self-improving AI systems, while Periodic Labs, also backed by Andreessen Horowitz, raised $200 million to apply AI to materials science discovery.
Mirendil is pitching the underlying research automation infrastructure that could sit beneath both. That is a harder thesis to validate, and a larger prize if it holds.
Investors and plans for the capital
Andreessen Horowitz and Kleiner Perkins have backed some of the most heavily funded AI companies of the past two years. A16z led Thinking Machines Lab’s seed round and Periodic Labs’ $200 million Series A; Kleiner Perkins co-led Harvey’s $300 million Series E in 2025 and led Hippocratic AI’s Series B.
Their joint entry into Mirendil’s seed signals a shared conviction that automating AI research is the next structural bet in the space. TO COME: founder quote on plans for the capital and why these investors.]
If AI can meaningfully automate its own research cycle, the advantages that today’s frontier labs hold, thousands of researchers, proprietary infrastructure, and years of institutional knowledge, become less decisive. That is exactly what Mirendil is betting on. Whether the timing is right, or five years early, is what the $200 million is there to find out.
