◉CLAUDE LABJP
●SUNSET — The legacy Workbench and experimental prompt tool APIs retire tomorrow, August 17, one day out●LATEST — Version 2.1.233, released August 15, is current: GitLab merge request URLs now work with --worktree and the claude agents view, where MRs appear as !N●SECURITY — Windows paths written with the NT \??\ device prefix no longer bypass UNC validation, closing an NTLM credential-leak vector●TODO — Todo and task tracking tools are off by default on Opus 4.8, Sonnet 5, Fable 5, Mythos 5, and newer models; set CLAUDE_CODE_ENABLE_TODO_TOOLS=1 to bring them back●FORK — Version 2.1.232, released August 14, makes subagent_type: 'fork' the default, so a forked subagent inherits the full conversation and prompt cache●MENTION — Typing @ in the prompt now mentions another Claude session by name, and SendMessage reaches that session directly●SUNSET — The legacy Workbench and experimental prompt tool APIs retire tomorrow, August 17, one day out●LATEST — Version 2.1.233, released August 15, is current: GitLab merge request URLs now work with --worktree and the claude agents view, where MRs appear as !N●SECURITY — Windows paths written with the NT \??\ device prefix no longer bypass UNC validation, closing an NTLM credential-leak vector●TODO — Todo and task tracking tools are off by default on Opus 4.8, Sonnet 5, Fable 5, Mythos 5, and newer models; set CLAUDE_CODE_ENABLE_TODO_TOOLS=1 to bring them back●FORK — Version 2.1.232, released August 14, makes subagent_type: 'fork' the default, so a forked subagent inherits the full conversation and prompt cache●MENTION — Typing @ in the prompt now mentions another Claude session by name, and SendMessage reaches that session directly
TAG

claude-opus-4-8

1 articles
← Back to all tags
Related:
prompt-caching1claude-sonnet-51model-migration1cost-optimization1
⬡ API & SDK/2026-07-02Advanced

Your Cache Hit Rate Resets to Zero the Morning You Switch Models — Prompt Cache Rewarm Design for the Opus 4.8 to Sonnet 5 Cutover

Prompt caches are scoped per model, so day one of a model migration starts at a 0% hit rate. Why percentage-based rollouts break cache economics twice over, and how cohort cutover by task family preserves them — with working measurement code.