Austin Griffith(@austingriffith):We tested if ethskills actually helps AI models get Ethereum right. 64 evals. 2 models. With and without skill docs loaded. Results: • Claude Opus 4.6: 33% → 95% • GPT-5.4: 16% → 95% Thanks to @AnthropicAI for the idea, @escottalexander and @_carletex

Austin Griffith

@austingriffith

🛠🔥🧙‍♂️ I work for Ethereum at the EF 👉 A human should learn how to build at 🤖 An AI agent should use

加入 March 2007

10K 正在關注 69.7K 粉絲

Austin Griffith@austingriffith

2026.03.10 21:54

We tested if ethskills actually helps AI models get Ethereum right. 64 evals. 2 models. With and without skill docs loaded. Results: • Claude Opus 4.6: 33% → 95% • GPT-5.4: 16% → 95% Thanks to @AnthropicAI for the idea, @escottalexander and @_carletex_ for surfacing it.

736

轉發到社區

熱門用戶