Editorial contract

Evidence before certainty

Automation may discover a source and prepare a draft. It cannot promote that draft into a confirmed fact without explicit editorial approval.

Four public labels

Confirmed

Confirmed

An official source or directly observed outcome supports the specific claim.

Signal

Signal

A first-party statement matters, but the reset outcome or full scope remains unverified.

Editor's forecast

Editor’s forecast

An exact probability inside a bounded window, with evidence class, rationale, falsification condition and revision time.

Community forecast

Community forecast

An anonymous aggregate that can reveal expectations but cannot establish a fact.

Source priority

  1. OpenAI status, documentation and release notes.
  2. First-party statements from relevant OpenAI leaders or builders.
  3. Directly verifiable product behavior with a dated scope.
  4. Community reports as discovery inputs, never sole confirmation.

Probability ceilings

The prominent percentage means the estimated chance that a global reset happens inside the displayed window. It is not source confidence and it is not the community vote.

Up to 49%
General limits discussion without a reset-specific statement.
Up to 79%
A credible first-party hint that specifically mentions a reset.
Up to 95%
A direct first-party signal that a reset is likely or imminent.
Up to 99%
A near-explicit timing notice while the event is still unobserved.

Once the reset occurs, the state becomes Confirmed—never a 100% forecast. These editorial probabilities are not statistically calibrated at launch; resolved history will make later accuracy reporting possible.

Corrections and lifecycle

Published records are versioned. Corrections append a dated revision; retractions remain visible in history rather than disappearing. Signals may become confirmed, resolved, corrected or retracted.