--- id: "2026-10-21-mirendil-base-one-researcher-agents-pretraining" url: "https://postcutoff.com/e/2026-10-21-mirendil-base-one-researcher-agents-pretraining/" as_of: "2026-11-19T22:11:00+01:00" date: "2026-10-21" date_precision: day category: research importance: 3 confidence: medium status: [Partly confirmed] sources: 2 editor: human editor human_review: null version: "2026-10-27" --- As of: 2026-11-19 22:11 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-10-21-mirendil-base-one-researcher-agents-pretraining/ # Mirendil-Base: one researcher, agents, <$10M Full title: Mirendil says one researcher with AI agents pretrained a frontier-scale coding base model in 3.5 months for under $10M On Oct 21, 2026 Mirendil, the ex-Anthropic startup working on self-improving AI, announced Mirendil-Base, a 100B–500B-parameter coding base model. It says a single researcher built it in 3.5 months for under $10M in total compute, with AI agents writing over 99% of the code and running and debugging the experiments. On bits-per-byte over seven outside code repositories it beats DeepSeek-V4-Pro-Base, GLM-5.3 and Nemotron-550B-Base and trails the post-trained Kimi K3. Larger Large and XL versions are still training. All of this is self-reported, and no weights, API or paper have been released. ## Key facts - Family: Mirendil-Base (100B–500B parameters), Base-Large (500B–2T, ~80% through training) and Base-XL (2T–10T, ~19%), described as 'coding-specialized models'. Post-training has started - Cost and team: under $10M all-in (agent tokens, data processing, research, small experiments and the final run) over 3.5 months, with one unnamed researcher. Agents wrote over 99% of the code; the researcher chose what to try and what to trust - Data: only public data (GitHub and other public code, web pages, public datasets); no purchased or licensed data, no distillation and no synthetic data. Code written from scratch on PyTorch and JAX - Bits per byte on seven external repositories (lower is better): Kimi-K3 (post-trained) 0.193, Mirendil-Base 0.222, GLM-5.3 0.242, DeepSeek-V4-Pro-Base 0.243, Nemotron-550B-Base 0.247. Mirendil says Large and XL both already exceed Kimi-K3 (as summarised by Import AI) - Mirendil: 'We don't think we have reached fully autonomous recursive self-improvement yet, but we are closer than we have ever been.' Jack Clark's Import AI 478 (Oct 26) headlined it 'Mirendil shows what RSI means' ## What happened Mirendil says it pretrained a large coding base model without a pretraining team: one researcher directed AI agents that wrote the training stack, ran the experiments and diagnosed the failures. Its only evidence is a compression metric (bits per byte) on outside code. No downstream benchmark of a finished model, no weights and no technical report have been published. ## Why it matters Pretraining a base model of this class used to take large teams and nine-figure budgets. If the claim holds, agents are already compressing frontier-lab R&D into one person's work, which is the early stage of AI-driven AI research that labs and forecasters watch for. It is still a self-reported result from a startup that is raising money. ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 143 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 113 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 204 days after its cutoff - Grok 4.7 (training cutoff May 2026): 143 days after its cutoff ## Sources 1. [Mirendil: Pretraining at scale without a pretraining team](https://mirendil.com/news/pretraining-at-scale-without-a-pretraining-team/) (mirendil.com, official) 2. [Import AI 478: Math for alignment; ProgramBench; and Mirendil shows what RSI means](https://importai.substack.com/p/import-ai-478-math-for-alignment) (importai.substack.com, discussion) ## Changes - 2026-10-26 (filed): Created (Mirendil post read directly; five days after publication) ## Related - 2026-09-22: [Mirendil, an ex-Anthropic startup building self-improving AI, in talks to raise up to $1B at a $5B valuation](https://postcutoff.com/e/2026-09-22-mirendil-5b-valuation-talks/index.md) - People: [Jack Clark](https://postcutoff.com/person/jack-clark/)