Anthropic reportedly improved Claude by stripping down its system prompt
jasonkneen · x · 2026-07-25
Apparently Anthropic stripped a large amount of instruction text out of Claude’s system prompt and saw a big improvement, suggesting the models may need far less steering than people assume.
The post argues this should be a reminder for builders: before adding more tools, prompts, and guardrails, try removing parts first and see whether the model performs better with less guidance.
Related event: Anthropic Slashes Claude Code System Prompts by Over 80%(11 posts)→
More from Models
- Aidan Gomez: model's CoT said no tool calls right before making them — aidangomez · 2026-10-27
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27
- ChatGPT's crawler sample code strips out all Schema markup — VeryWellVersed · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27