Claude 5

1. Claude 5 family's hallucinations look a lot like internal Anthropic emails (github.com via hn)

   +11 7w anthropic

   The script used to gather, and the outputs related to, this blog post.

2. Audit found an unused skill and took action. It was still costing me 8,800 tokens every message for months after. (www.reddit.comhttps)

   7w anthropic

   Anthropic published new context-engineering rules for the Claude 5 models. I read it and did the obvious thing: pointed Claude at my own repo and asked what it was costing me before I typed a single word.

← all threads