Victoria Bazir
⬤ Former researchers from Anthropic are launching a new AI startup focused on scientific discovery. The team is in talks to raise around $175 million at a valuation of roughly $1 billion for a company called Mirendil. Its systems are designed to accelerate breakthroughs in fields like biology and materials science.
⬤ Mirendil is led by former Anthropic researchers Behnam Neyshabur and Harsh Mehta. The company plans to build AI models capable of long-term scientific reasoning, helping researchers explore complex hypotheses and run computational experiments. This fits into a broader industry push toward so-called "neolabs" built specifically for scientific work.
Autonomous AI researchers capable of generating real scientific breakthroughs could emerge within the next decade.
⬤ Venture firms Andreessen Horowitz and Kleiner Perkins are reportedly in talks to co-lead the funding round. If closed, the deal would place Mirendil among a new wave of AI startups applying large-scale models to research challenges, a trend explored in the AI efficiency race between OpenAI and Anthropic models.
⬤ Interest in AI-driven discovery is growing fast. OpenAI CEO Sam Altman has suggested autonomous AI researchers could emerge within the next decade. Startups like Mirendil aim to build systems for sustained reasoning and experimentation, and efficiency work like the TAPPA framework for reducing LLM costs is making such systems more practical. Meanwhile, tools like Anthropic Claude Code review tools show how AI is already reshaping technical workflows across industries.
Victoria Bazir
