Google just put agents in charge of model post-training. autofinetune, out september 11, uses ai…
google just put agents in charge of model post-training. autofinetune, out september 11, uses ai agents to tune hyperparameters and optimize llm training loops on tpus.
the tuning run stopped needing a human.
Context
Google's developers blog post of September 11, 2026, Autonomous LLM post-training with Tunix on TPUs, describes autofinetune as applying autonomous research loops to LLM post-training, inspired by the autoresearch project and built with Tunix on TPUs. A human writes a Markdown spec that defines the loop, boundary conditions, evaluation criteria and constraints. The agent then edits the training script and hyperparameters, runs the job, measures the target metric, keeps winning commits or reverts regressions, and logs results. It adjusts LoRA rank and alpha, the optimizer, the learning rate and similar settings. The first case study is supervised fine-tuning of google/functiongemma-270m-it on google/mobile-actions, with function call accuracy as the metric.
A human sets the objective, boundaries and metric, so agents in charge is supportable only in that bounded sense, and the tuning run stopped needing a human is the author's take. The post reads as an example setup, and no first-party product label or availability tier was found, so product status is unverified. The accuracy gain from the case study was not extracted and is not quoted. A separate AlphaEvolve example repository was seen as a snippet and is not autofinetune.
Watch next
- A released repository and license, and the second case study's results.
Sources
- Google Developers Blog: autonomous LLM post-training with Tunix on TPUs (September 11, 2026)developers.googleblog.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 23 September 2026 at 02:04 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →