A product in the world. A place to test what comes next. A creative practice that keeps making.
A dated account, not a live status board. Published work, synthetic rehearsals, and historical experiments remain distinct.
01 / One body of work, several places
Different jobs. Shared continuity.
01 / WindowsMake & rehearse
Synthetic operation
A service made it through the cycle.
A five-item fictional service went through review, approval, publication, live progression, a corrective step back, and completion. Its recorded history could be reopened.
Evidence & limits
The September 19 rehearsal retained 12 actions and bound the result to the exact synthetic candidate. The local Forge bridge also retained bounded job state across a restart.
Not established: Scripted candidate proof is not an unaided church-operator rehearsal or continuous unattended operation.
The ChurchOS Evolution Museum and the Forge sandbox responded during the evening inspection. Recovered product workflows sit alongside new interface studies.
Evidence & limits
The retained museum publication receipt records 19 cards, 24 recovered objects, and 66 successful curated-link checks. Receipt integrity was checked; the link suite was not rerun for this edition.
Not established: These are private research surfaces. A responding sandbox does not establish that every tool works or that an autonomous worker is operating.
An iPhone simulator displayed the synthetic service with its five program items and the print, Community, Rooms, and QR controls.
Evidence & limits
The public program was visually inspected in an iPhone simulator on the Mac, with a preserved screenshot.
Not established: This does not prove physical iPhone behavior, authenticated roles, offline recovery, or that every visible control completed its workflow.
The field Windows workstation supplies the real setting: a church operator, a browser, and the output equipment around them. That environment is separate from the Lab.
Evidence & limits
The field contribution reports ChurchOS reachable in Edge. Windows production checks also observed signed-in navigation and service-bound room re-entry on the current public source.
Not established: Current-candidate printing, the sanctuary display chain, authenticated Finance replay, and a real service remain separate gates. No complete physical rehearsal is claimed.
The Community walkthrough exposed a broken continuation: a service-bound Rooms link lost its context at sign-in. A bounded repair restored the route back to the right service.
Evidence & limits
The September 19 record includes candidate checks, promotion, and an authenticated production read-and-reopen journey. Production source and the older release label were recorded separately.
Not established: That production repair is not the exact frozen Build 50 rehearsal candidate. Neither proves full Sunday readiness.
The product walk began with what already existed. Community, programs, and private Rooms were real capabilities; the useful change was joining the journey between them. The public service also serves a revision-bound FSBFMC folded bulletin.
Observed production read/reopen behavior. Full operational writes and physical output remain separate proofs.
02
Memory a new worker can use.
The archive work connects retained sources to present questions. The recorded 21-concept query sweep completed without query failures; a bounded Homebase query returned decisions, implementations, unresolved work, evidence, and later changes together.
Retrieval works within the inspected coverage. Contradictions, missing sources, and historical authority are not resolved by returning a summary.
Local candidate commissions retained the sequence from acceptance to artifact and verification. A restart recovered the same job state. The first failed provenance check was kept, and the persisted-byte hashing was corrected.
Bounded job continuity—not a completed 24–72-hour autonomous operating window, and not a qualification of every local-model task.
The historical-math baseline records what can be measured and what is missing. A small context pilot suggested less inference work; it did not establish general savings. Model loading, acceptance, repeated runs, and source coverage still matter.
Nine selected historical observations are not a company-wide productivity denominator. No dollar-saving claim is made.
03 / Studio, with something inside it
Not everything needs to be useful.
Origin / After Hours was already in the collection. This edition gives it a door: a visual instrument, procedural posters, and a short film. Change the composition. Keep an editable preset. Make something because you want to.
TCO has not declared the entire operating company proved. Real Sunday operation, the complete continuity window, physical Apple and display behavior, and unresolved archive relationships remain open.
The useful change is that those limits can be named. An experiment can succeed without pretending to be a product. A product can improve without pretending the work is finished.
