Known limitations & next work
The remaining work is explicit rather than hidden behind catalogue totals.
Complete earned raider, professional-main and solo-generalist Floor 1 journeys have not been demonstrated. The current exact-replay journeys exercise important mechanics but cover a smaller scope.
Observed multiplayer sessions, long-lived market populations, Discord desktop/mobile rehearsals, measured latency and large-load behavior remain to be collected. The engine currently serializes a single SQLite world aggregate; history retention and contention need production planning.
The laboratory reports quality ceilings in hunting, medicine, mining, tailoring, tanning, tracking and weaving, plus field shortages and unfinished durable goals under its synthetic assumptions. These are design review findings, not proof that players cannot complete those careers.
Salvage is a known current defect: the recipe selector expects an id inside each recipe record, but installed recipe IDs are the mapping keys. The action raises an error before completing; its registered form is not evidence of a working transaction.
The authored text still has an editorial queue. Narrative validation, broad accessibility review, advanced policy editing, full recurring-content scope, post-clear longevity and full first-floor acceptance are unfinished.
The public documentation is deployed through private object storage and a CDN. Agent workers are supervised on the current host. Off-host agent backups, a host-loss recovery exercise, broader game service migration, population monitoring and a public authenticated game frontend remain separate operational work.
There is no live LLM game master that invents rewards, changes canon, or edits a running world. The specialist agents inspect bounded source packets and produce reviewable drafts and findings.