The sun is being tamed by reinforcement learning
the sun is being tamed by reinforcement learning: princeton’s pacman framework controls fusion plasma in milliseconds and predicted a damaging instability 200 milliseconds ahead on the diii-d tokamak, while deepmind drove the tcv reactor in real time. iter’s ai surrogates run physics simulations up to a million times faster.
the reactor’s heartbeat is now a model.
Context
The Princeton Plasma Physics Laboratory announced on Sep 2, 2026 PACMAN, an AI framework for controlling fusion systems that makes key decisions in milliseconds. Its design paper describes Prediction And Control using MAchiNe learning on the DIII-D tokamak.
Earlier work in Nature (Feb 21, 2024) used deep reinforcement learning to avoid plasma tearing instability. DeepMind has a public library of surrogate transport models for tokamak fusion, and an arXiv paper covers a fast surrogate of the LIUQE reconstruction for TCV.
PACMAN's millisecond decisions on DIII-D and the DeepMind and TCV work match their sources. The 200 millisecond early prediction of an instability, Princeton using reinforcement learning for PACMAN and ITER surrogates running up to a million times faster are not seen in the sources read, so unsupported here, not refuted. The Nature RL paper is a 2024 related study.
Related work
Watch next
- Read the PPPL release for how far ahead PACMAN predicts.
Sources
- PPPL, Sep 2, 2026pppl.gov
- Nature, Feb 21, 2024nature.com
- google-deepmind/fusion_surrogates on GitHubgithub.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 2 October 2026 at 08:59 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 →