Tags

Tag: Prompt engineering

8 posts

Sunday11 October

Anthropic18:09
Claude Code’s system prompt shrank by over 80%, scores held
Claude Code would get “leave documentation as appropriate” and “DO NOT add comments” in one request. Anthropic’s fix was to cut over 80% of its system prompt, with no measurable loss. Would you strip your own Claude rules that hard, or keep the safety net?

Friday2 October

OpenAI20:06
OpenAI’s GPT-6 guide says drop blanket “always ask” rules
Teams running GPT-6 agents on jobs that last hours or days can send a correction partway through. OpenAI’s guide says the correction waits in a queue and doesn’t stop tools that are already running.

Friday25 September

Benchmarks00:13
Claude Fable 5.1 kept writing comments it was told to skip
Told not to add comments, Claude Fable 5.1 still wrote new ones in 33 of 100 SWE-bench Verified tasks. GPT-6 Astra did it in 6%, though it resolved fewer issues.

Wednesday23 September

Anthropic03:22
Anthropic tells Opus 5.5 users to delete “think hard” lines

Thursday10 September

Anthropic19:09
Prompt audit in Claude Code cuts cost 14.6%

Thursday3 September

Anthropic22:04
New Claude prompt refuses lyrics and Sonic banners

Thursday27 August

Anthropic08:14
80% of Claude Code’s system prompt got cut

Wednesday12 August

Research11:14
Your CLAUDE.md gains 4.9 instructions per commit and never loses them