🌏 中文版
Funding Details
| Item | Value |
|---|---|
| Company | Mirendil (United States, San Francisco) |
| Round | Series A |
| Amount | $1B |
| Lead | Kleiner Perkins, Clearlake Capital |
| Participants | Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC, S32, and others |
| Valuation | Not formally disclosed (Bloomberg reported in September that talks were targeting $5B; the final figure may differ) |
| Total raised | $1.2B ($200M seed + $1B Series A) |
| Founded | Early 2026 |
| Employees | ~20 at the seed stage, drawn largely from Anthropic, xAI, Google DeepMind, and OpenAI |
What the company does
Mirendil is building AI agents that automate frontier AI research itself — the pitch is that a single researcher paired with a fleet of agents can do what used to take an entire team.
Over the past three weeks the company has shared its first public results, all pointing the same direction: it pretrained Mirendil-Base, a competitive coding model, from scratch in 3.5 months for under $10M in compute, then post-trained it into Fern, which holds up to 10 million tokens of context — enough for an agent to work continuously for days. Lor, now in private beta, lets hundreds of agents write code and run experiments in parallel for days at a time; Mirendil itself runs its own research on Lor every day.
Co-founders Behnam Neyshabur and Harsh Mehta met at Google in 2019, joined Anthropic together in late 2024, and left in December 2025 — their argument being that the capability to do frontier AI research shouldn't be confined to a handful of large labs. The company remains pre-product, and investors were already discussing a $5B valuation just three months in.
What this round signals
What it means for the agent ecosystem
This round signals that agents aren't just being shipped as products — they're being used to automate the AI research process itself. Mirendil's pitch isn't another chatbot or coding assistant; it's "replace a research team's headcount with agents," compressing what used to take a full lab down to a handful of people plus agents.
- June 2026: $200M seed
- September 2026: talks reported
- November 2026: $1B Series A
Simplified. The three data points come from different sources (an official announcement, a Bloomberg report, and an official announcement that withheld valuation) with varying precision.
What the investors are betting on
Kleiner Perkins and Clearlake Capital already led the seed round (June 2026, $200M at $1B) and doubled down for this Series A — a sign they're not betting on a single deal's return but on the broader thesis that "democratizing frontier AI R&D" will keep attracting capital. Less than a year after the founders left their jobs, reported valuation talk has already risen at least 5x, showing top-tier VCs are willing to pay for pedigree and early signal alone, without waiting for product or revenue proof.
Numbers worth watching
- Only about five months separate the seed round from the Series A (June to November 2026) — far faster than the typical AI-startup fundraising cycle
- Total disclosed funding is $1.2B, yet the company still has no public product roadmap or revenue figures — this is purely a "trust the team" valuation model for now
- Bloomberg's September report put the negotiating target at $5B; if the final number lands near that, the valuation rose 5x in five months
Watchlist status
Mirendil is not yet tracked in the watchlist. We'd recommend adding it to section A2 (AI models/platforms) — it's a leading example of commercializing "using AI to do AI research" with agents, and it's worth tracking whether its models (Mirendil-Base, Fern) and agent platform (Lor) eventually open up to outside customers.
Today's takeaway
A company less than a year old, with no shipped product and no revenue, just raised $1B — a sign we're in a stage where pedigree and narrative are worth more than product. But the twist is that investors are betting agents themselves will make that "small team, huge ambition" story real: the money that used to require proven product-market fit now only needs early signal that "agents can actually replace a research team."
References
- Mirendil Raises $1B to Democratize Frontier AI R&D
- Mirendil raises $1B to put frontier AI research within reach of small teams
- Mirendil Raises $1 Billion Series A for Agent-Led AI Research Tools
- Report: Ex-Anthropic duo's Mirendil in talks for $5B valuation just 3 months after $1B seed
- Ex-Anthropic Staffers' Self-Improving AI Startup in Talks to Raise at $5 Billion Value