Neo(@NeoAIForecast):GPT-6 Astra just debugged another AI by reading its weights. MIT CSAIL pointed GPT-6 Astra at Qwen3.5-0.8B, a model Qwen's own card says gets stuck in thinking loops. It traced the loops to the MLPs in layers 15 and 19 and edited 0.016% of the weights. MATH-500: 7.9% to 48.1% GSM8K: 13.8% to 67.1% IFEval: 21.0% to 62.3% The real story behind this is you can only patch weights you can download!!! Open weights win!!!

2026.11.02 22:03

GPT-6 Astra just debugged another AI by reading its weights. MIT CSAIL pointed GPT-6 Astra at Qwen3.5-0.8B, a model Qwen's own card says gets stuck in thinking loops. It traced the loops to the MLPs in layers 15 and 19 and edited 0.016% of the weights. MATH-500: 7.9% to 48.1% GSM8K: 13.8% to 67.1% IFEval: 21.0% to 62.3% The real story behind this is you can only patch weights you can download!!! Open weights win!!!

Rikiya Takehi@rikiyatakehi

2026.11.02 17:51

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: