Simon Willison gave a large model, Qwen 3.8 27B, the raw weight numbers of small game-playing neural networks and asked which game each plays; it named 3 of 4

There's a buzzy new paper about GPT-6 Astra being able to directly interpret model weights (prompts full of floating point numbers) - I tried it with Qwen 3.8 27B on my laptop and found that, after 45+ minutes, that model could achieve the same thing simonwillison.net/2026/Nov/3/qw…