Under the hoodTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
Prompt CacheArea: what it touches
Internal ChangesKind: in v2.1.288,
Internal ChangesSection of the release
What
Claude Code reuses parts of earlier requests to the model through a prompt cache, which stores a prompt's opening so it does not have to be processed again. A cache breakpoint is a marked point in the prompt up to which the cache can be reused.
Claude Code can now mark the end of a parent prompt, meaning the prompt that started a piece of work, as one of these breakpoints. This is controlled by an optional setting inside Claude Code. The usage data Claude Code records about cache breakpoints now also notes where a request came from and whether the parent prompt's end was marked.
Why
Where breakpoints sit decides how much of a prompt can be reused from the cache, so this can change how often requests are served from it.
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtIt is not clear which parts of Claude Code turn this marking on.
