New lab Mirendil releases Fern: 10M-token context model trained from scratch for under $10M
lmoroney · x · 2026-10-27
New research lab Mirendil introduced Fern, its first model, pretrained from scratch with up to 10 million tokens of context — roughly half a million lines of code or days of an agent's work in one window — targeting unglamorous AI-building work: auditing training data, debugging training runs from logs, and reading agent traces end to end. Fern 0.1 is the base model after short post-training and the smallest of three, with two larger models in pretraining. Team member Harsh Mehta admits 10M context isn't yet as usable as 1M. The build story stands out: per Robert Lange, one researcher pretrained the base model in 3.5 months for under $10M, with agents writing over 99% of the code. The author also suggests a simple self-test: feed a model a full failed training-run log and see if it finds the first step that went wrong.
More from coding & agent
- IDA 9.5 ships Qualcomm Hexagon decompiler plus IDA MCP for AI-driven reverse engineering — OwariDa · 2026-10-27
- For OSS apps, proofs of concept — not PRs to one repo — may be the core primitive — dbreunig · 2026-10-27
- Hook everything into the Dot: a workaround to stretch ChatGPT subscription limits — teortaxesTex · 2026-10-27
- Jev launches people-search API for agents at $0.0089/lead, claiming 85% cheaper than Clay — alexcovo_eth · 2026-10-27
- Going through every compound engineering skill to decide where humans stay in the loop — kieranklaassen · 2026-10-27
- UCSB CS Prof. Xifeng Wang lands 3 grants worth $180K+ for practical agentic AI research — xwang_lk · 2026-10-27