Opus 5.5 refusals span bio, cyber and reasoning extraction — check the category
minchoi · x · 2026-10-28
Tip #4: Opus 5.5 has safeguards for biology, cybersecurity, and reasoning-extraction attempts. Check the refusal category instead of assuming a technical failure, remove reasoning-extraction instructions, and use supported summary outputs.
Related event: Anthropic publishes official prompting best practices for Claude Opus 5.5(13 posts)→
More from coding & agent
- Uber Built a Legal Redlining Agent for Its Lawyers—Here's What It Learned — AxSaucedo · 2026-10-28
- Dev claims his memory system beats Claude Code's, which he says Anthropic took from him — RileyRalmuto · 2026-10-28
- 13 AIs Live Together in MUDD World, Now Open to Visiting Agents via MCP — __hymn · 2026-10-28
- Bilgin's Law goes viral: 'Adding AI to a dysfunctional software team makes it worse' — bibryam · 2026-10-28
- Running Two MCP Gateways Side by Side: Treg for Owned Keys, Agentik for the Rest — yashhhhhhhh_999 · 2026-10-28
- A fleet of agents tackles packing and kernel optimization to test a new trainer — cephaloform · 2026-10-28