Anthropic’s most powerful AI model got banned by the US government 3 days after launch.
3 weeks later, it’s back — with a classifier that blocks the jailbreak that got it pulled 99% of the time.
First frontier AI model ever pulled by government order.
@AnthropicAI#ClaudeAI #AI
Writing Anthropic prompt-caching boilerplate sucks, so I built a zero-config wrapper that handles it for you — and prints your cache hit-rate + $ saved right in the terminal.
npm install claude-cache-master
before:
system: [{ type: "text", text: docs, cache_control: { type: "ephemeral" } }]
after:
client.createWithCache({ systemDocs: docs, messages })
It also warns you if you accidentally put a timestamp/UUID inside a cached block (silent killer of cache hit rate).
https://t.co/KlSfeHVpnd
#ClaudeAI #buildinpublic