A Reason to Wake

A reason for the next turn

The bounded core now waits for the initial briefing or a relevant shared event, such as completed work or an incoming message, before taking another planning turn. Time passing alone does not wake it. Its budget, cooldown and consumed events survive a save. It can still see grounded opportunities when it plans, but changing opportunities alone are not a trigger. Private pawn needs and thoughts stay private.

Pawns also gain a not-now reply. Deferring retires the offer without a job or a promise; this bounded planner suppresses further ordinary offers to that pawn for the session. Automatic after-meal resumption is not implemented. A legitimate later invitation will still need fresh consent.

Rehearsal and live choices

In scripted rehearsal, Alvin deferred. Beatrice countered with a smaller haul, received a revised offer, then freshly accepted and delivered five wood. Completion triggered a wait. This tested the mechanics, not a live preference.

In the live trial, Alvin accepted two separate offers. Three trips delivered thirty wood from the first stack; that completion triggered a fresh offer on a different stack. Three more trips delivered thirty more wood, then completion triggered a wait. The core correctly distinguished the source IDs. Passing time produced no fourth turn.

Three core and two pawn calls were used, below the four-and-five ceilings, with no Jev calls or rerolls. All 102 sampled native-activity states were unpaused across roughly two minutes; inference was explicitly paused. No live deferral or question was chosen, so this does not show reduced agreeableness or verify the hunger-testimony wording fix in live use.

The unfinished record

The last topic prose called both hauls resolved, but the stored topic labels remained open and deferred. Completion receipts were correct; topic bookkeeping was not. Paired and full cold restore preserved both the results and that mismatch without extra calls. The saved result was not rerun to tidy it.

All 272 Node tests and ten Python transport checks passed. Independent review found missed movement-completion events; the fix passed regression tests and focused review. Next are explicit topic lifecycle and a legitimate route back from not-now. Construction and cooking remain separate work. This is one authored fixture, not general strategy, and its changed timing prevents a controlled comparison with the earlier fixed-round trial.