Report: Ex-Anthropic duo’s Mirendil in talks for $5B valuation just 3 months after $1B seed
Tech Funding News Abhinaya Prabhu ● Covered by 3 sources
Mirendil is talking a $5B valuation just 3 months after its seed. The AI lab still has no public product, but big investors keep betting anyway.
Based on reporting by Tech Funding News, Abhinaya Prabhu — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Mirendil is in talks for a $5 billion valuation, five times its level in June, even though the San Francisco startup still has nothing public to show. Kleiner Perkins may lead a round of up to $1 billion, and Andreessen Horowitz is also said to be negotiating another investment. Neither firm has commented publicly, but the appetite is obvious: money keeps showing up before a product does.
The company was founded by four former staffers from Anthropic, xAI and OpenAI, with Behnam Neyshabur and Harsh Mehta at the center of it. Neyshabur and Mehta met at Google in 2019 after Mehta cold-emailed him. Neyshabur later co-led Google DeepMind’s Blueshift team, working on reasoning and math for Gemini, while Mehta built Anthropic’s internal AI-research platform, initially as a team of one. They left Anthropic together in December 2025, taking Shayan Salehian and Tara Rezaei Kheirkhah with them.
Mirendil’s pitch is unusually specific for a category that often sounds like pure vapor. It says it will train models that are especially good at AI research, then wrap them in a system that plans experiments, runs them, and keeps iterating with less human input. The idea is that a biology lab or materials-science team could get frontier-style research tooling without building a frontier lab first. That’s the business model, or at least the hope.
The bigger point is that Mirendil is riding a market already willing to pay absurd prices for research automation. Safe Superintelligence has raised roughly $3 billion in disclosed equity at a $32 billion valuation, plus a separate $5 billion Nvidia compute deal, with no product. Recursive Superintelligence raised $650 million at $4.65 billion. Thinking Machines Lab is now back in talks near $40 billion, and at least it has shipped Tinker and Inkling. In that crowd, Mirendil’s main asset is not a launch. It’s pedigree and timing.
My take — AI-written commentary, not fact-checked reporting
This is the kind of deal that makes AI look less like a product market and more like a talent auction with a very expensive coat of paint. Investors keep paying for the promise that labs can automate their own R&D, which is fine until everyone realizes the demo is still a funding deck. The pattern is getting familiar: if there’s no product, just call it a platform and add a few zeroes.
Read more about this at: Tech Funding News