GPT-6 editing Stockfish NNUE weights shows +8.1 Elo gain, pending proper testing
imjustnewatai · x · 2026-11-10
A community experiment update: GPT-6 Astra editing Stockfish's NNUE weights produced a best build with +8.1 Elo gain, though with a wide confidence interval (-14.9 to +31.3 Elo), not yet statistically significant.
Author imjustnewatai says the next step is putting the build through the "fish test" (standard engine match verification). Using an LLM to directly rewrite evaluation network weights and measure Elo is an unusual, counterintuitive experiment — result still unproven.
Related event: 78 Astra agents spend 3 days tuning Stockfish weights for ~11 Elo gain(4 posts)→
More from Fun
- Ethan Mollick had Fable 5.2 build a full reverse Scooby Doo game with its own trailer — emollick · 2026-11-10
- Joke video shows Jensen Huang 'escaping his handlers' en route to Megaspeed QBR — julsimon · 2026-11-10
- Vibe-codevember day 8: WebGPU tesseract built with Gemini 4 and Antigravity — AI_Andrew · 2026-11-10
- Fable 5.2 builds a playable voxel engine from a single prompt — mattshumer_ · 2026-11-10
- Non-native English speaker mocked for using ChatGPT to write Reddit posts — Naive-Consequence110 · 2026-11-10
- ChrisGPT claims he leaked Fable 5.2's '30% faster' claim weeks before Anthropic's announcement — ChrisGPT · 2026-11-10