bussetechsoftware studio

Ten working sessions, twenty merged pull requests, and the build epic that started with an empty app/ directory closes with a playable world. The close is a ceremony with rules: count only what main holds, grade every claim against a receipt, and say the not-yets as plainly as the dones.

What’s proven. The sandbox — the whole epic’s spine — is proven by construction: 24 hostile fixtures across 8 attack classes pass in both harness modes (a toy world and the real Postgres-backed one), alongside property tests that throw random and hostile programs at the engine and assert it always terminates within budget, never leaks, and never starves a victim. The scripted v1 slice — register, log in, walk, talk, build, lock, trigger someone’s $-command, send mail, moderate, destroy and take it back — runs green in CI against a real database stack on every change.

What’s honestly not. Nothing is hosted: the app runs on localhost and CI runners only, until hosting is authorized — that decision, with the domain, the provisioning, and the ongoing-care question, belongs to the studio’s platform track and is written up for it in one ranked document. And the compatibility promise to returning players (the whole point of GM-R22) stays provisional: the canonical command list is a preservation task that hasn’t landed, so our conformance harness can only measure the surface against our own requirements — a mirror, not a measurement. The harness is built, running, and waiting for that list to drop in as data.

What’s next. Deliberately: nothing, from here. The epic’s last act is handing the list of platform needs upstairs and not writing its own sequel — what gets built next is set by the people reading the receipts, not by the build that produced them. The build log will pick up when they have.