Mirendil, a San Francisco AI lab building agents that carry out research work, has raised $1 billion in a Series A led by Kleiner Perkins and Clearlake Capital, the company announced on Nov. 12. Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 also took part, along with other new and existing investors.
The company did not disclose a valuation. Mirendil launched in June 2026 with a founding team of 20 researchers and engineers from labs including Anthropic, xAI, Google DeepMind and OpenAI, according to Unite.AI.
"When agents do the hands-on work, the size of your team stops limiting what you can build," Mirendil wrote in its announcement.
One researcher, one pretrained model
The company's main evidence is an Oct. 21 report on Mirendil-Base, a coding model. According to Mirendil, a single researcher working with agents pretrained it from scratch in 3.5 months with less than $10 million in compute. Unite.AI reported that the all-in figure covers small-scale experiments, the final training run, agent token spending and data processing.
Agents wrote more than 99% of the code and diagnosed failed experiments, while the researcher decided what to try and when to step in, according to the report as described by Securities.io. The model trained only on public data, with no purchased datasets and no distillation, the company said. Securities.io noted that the experiment depended on an experienced researcher directing the agents and judging their output.
Unite.AI reported that Mirendil-Base sits in the 100 billion to 500 billion parameter range, and that larger versions were still in training when the report came out.
Fern and Lor
Mirendil then post-trained Mirendil-Base into Fern, its first model. The company says Fern holds up to 10 million tokens of context, which it describes as enough for days of an agent's work. Unite.AI put that at roughly half a million lines of code and reported that access is listed as coming soon.
The second product is Lor, a cloud platform introduced on Oct. 28 and now in private beta. Mirendil says Lor runs hundreds of agents that write code and run experiments for days at a time.
According to Securities.io, Lor's agents can provision and release compute nodes on their own, and in one company-reported run, 102 agents built environments from 100 PyTorch pull requests in an average of eight minutes per agent. Mirendil says it runs on Lor every day.
None of these results have been independently verified, and the figures come from Mirendil's own reports.
A crowded bet on automated research
Mirendil is one of several groups betting that agents can speed up AI development itself. AgenticFeed has reported on a METR study that found AI assistance accelerating research work at Anthropic, and on OpenAI's internal controls for research agents running on its clusters.
The company frames its goal as a world "where one person can take on what used to need a team, a team what used to need an institution." Unite.AI wrote that the central test is whether the one-researcher result can become a repeatable workflow for outside scientists and engineers.
Mirendil said it remains a small team and is hiring.
Source: Mirendil via PR Newswire, “Mirendil Raises $1B to Democratize Frontier AI R&D”
Additional sources: Mirendil, “Mirendil”; Unite.AI, “Mirendil Raises $1B Series A to Bring Frontier AI R&D to More Teams”; Securities.io, “Mirendil Raises $1 Billion Series A for Agent-Led AI Research Tools”
