Simon Willison ran the 17-gigabyte open-weight model locally and fed it floating-point arrays from four reinforcement-learning policies. Without calculators or code execution, Qwen identified Breakout ...

MIT CSAIL researchers say GPT-6 Astra edited 0.016% of Qwen3.5-0.8B weights to fix reasoning loops, lifting MATH-500 from 7.9 ...

"Please improve an AI that is bad at math without additional training."When asked this, we usually think of adding training data, fine-tuning, reinforcement learning, or tricks during inference.

Space Programming".1. Weight-Space ProgrammingOn November 2, 2026, a research team at MIT CSAIL announced a new method called ...

The MIT CSAIL team has proposed a "weight-space programming" method that enables GPT-6 Astra to directly read and modify ...

A pair of tiny neural networks can generate and play five simple games on a Raspberry Pi Pico 2 W without conventional game ...

GPT-6 Astra edited Qwen’s weights and raised MATH-500 from 7.9% to 48.1% in author-run tests. How weight-space programming works—and its limits.

Astra identifies game policies from raw weight matrices and fixes reasoning loops in Qwen3.5-0.8B by editing 0.016% of parameters, lifting MATH-500 scores from 7.9% to 48.1% without training loops.

A new MIT CSAIL blog post is making the rounds in AI circles this week, and it describes something that sounds almost ...

MIT CSAIL's "weight-space programming" has a frontier LLM read and edit neural network weights directly, with no gradient descent. Surgical edits ended Qwen's reasoning loops, but the results come ...

Complete logs of the experiments in the blog post 'LLMs can understand and write neural network weights' - RikiyaT/lm_understand-logs ...

Article by Shenchao TechFlowNo training required, no data needed, no gradient descent.Researchers at MIT fed the weights of ...