Mirendil Raises $1B Series A for a Research Lab Where Agents Do the Hands-On Work
A billion dollars for a Series A, and the pitch is that the team size no longer matters. Mirendil announced the round on Thursday, led by Kleiner Perkins and Clearlake Capital, with Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 participating. The release gives no valuation, so neither do we. Release: https://www.prnewswire.com/news-releases/mirendil-raises-1b-to-democratize-frontier-ai-rd-302937248.html
The proof points are all about agents replacing headcount. Mirendil says one researcher, working with agents, pretrained Mirendil-Base, a competitive coding model, from scratch in 3.5 months on less than $10M of compute. That base was post-trained into Fern, which holds up to 10M tokens of context, which the company frames as enough for days of an agent's work. The third piece is Lor, now in private beta: hundreds of agents writing code and running experiments for days at a time. Mirendil says it runs on Lor itself every day.
What the money is for, in the company's words: a world where one person takes on what used to need a team, and a team what used to need an institution. In practice that means more compute and getting Lor out of private beta.
This is the cleanest 100X bet of the week. Most AI rounds fund a model or an app. This one funds the claim that frontier research itself can be run by a handful of people steering a swarm. The numbers are the company's own and nobody outside has reproduced the one-researcher pretraining run yet. If Lor's users can repeat it, the moat of the big labs, which is mostly a payroll of researchers, starts to look thinner.
← Back to all articles
The proof points are all about agents replacing headcount. Mirendil says one researcher, working with agents, pretrained Mirendil-Base, a competitive coding model, from scratch in 3.5 months on less than $10M of compute. That base was post-trained into Fern, which holds up to 10M tokens of context, which the company frames as enough for days of an agent's work. The third piece is Lor, now in private beta: hundreds of agents writing code and running experiments for days at a time. Mirendil says it runs on Lor itself every day.
What the money is for, in the company's words: a world where one person takes on what used to need a team, and a team what used to need an institution. In practice that means more compute and getting Lor out of private beta.
This is the cleanest 100X bet of the week. Most AI rounds fund a model or an app. This one funds the claim that frontier research itself can be run by a handful of people steering a swarm. The numbers are the company's own and nobody outside has reproduced the one-researcher pretraining run yet. If Lor's users can repeat it, the moat of the big labs, which is mostly a payroll of researchers, starts to look thinner.
Comments