budget: report how the compact slots are occupied #129
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "land/task/santi-slot-report"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
budget: report how the compact slots are occupied
Blocking on a full slot is only survivable if the occupancy can be seen
before the block arrives, and today it cannot: the audit reports an
estimate and a ceiling with no account of what the compacted part of the
timeline is made of. Report it beside the estimate, where the execution
envelope is already reported, from the same policy the prompt renders and
the enforcement will read. One computation, so a stated occupancy cannot
disagree with an enforced one.
Occupancy is not the count of compact records. A compact whose range lies
inside another's is never projected, because the assembly walks the
overlays in order and a wider one covers it, so the records outnumber the
slots that actually exist. Derive the spans from the message sequences
the boundaries name and keep only those no other span covers.
The first measurement against a real strand is worth recording: ten live
compacts against a ceiling of three, none covering another. That is the
shape of compacting the tail again and again without ever rewriting what
came before, which is exactly the habit the slots exist to interrupt.
The public schema gains two objects and a field, so the contract digest
moves with it. The document was diffed before the digest was touched: two
new schemas, one field, no path or removal.
Co-Authored-By: Claude Opus 5 (1M context) noreply@anthropic.com