Guides, news and explanations about AI and reliable systems.
Anthropic reports lower task costs at unchanged token rates. Existing clients must accommodate changed thinking settings, response blocks and forced-tool behavior.
Anthropic reports lower Opus task costs and new preserved-thinking behavior. A runbook exception provides a concrete comparison of the model’s reasoning and integration requirements.
What prompt-cache hits save once retries and failed work are included in task cost, with a local calculator and ledger.
Claude’s on-demand compaction lets an application replace older conversation history with a signed summary. A conditional restart shows why the summary must preserve relationships between instructions.
Claude Enterprise’s Compliance API beta now includes Chrome-session transcripts. Tool exchanges help locate an attempted action; the destination’s records establish whether it took effect.
A rollback exception tests whether Opus 4.6 preserves decisive evidence across a long session, and whether the resulting answer arrives in time to help.
