MIT study: GPT-6 Astra can directly rewrite the weights of other AI models
MIT researchers showed that OpenAI's GPT-6 Astra can read and deliberately rewrite the weights of other AI models. In one experiment, Astra raised a Qwen model's math accuracy from 7.9% to 48.1% by changing just 0.016% of its weights. The approach could fundamentally simplify AI optimization but still has clear limits.
