FORECAST TRACKER

Forecasts

English editions of official plans, reported roadmaps, analyst forecasts, and rumors, with later events used for verification.

From claim to outcome

  1. Claim recordedWho said what, when, and through which source.
  2. Waiting for evidenceExpected timing and public developments remain visible.
  3. Outcome assessedLater events support, partly fulfill, refute, or leave it unresolved.

Each English edition identifies its authorship and review state. The Chinese record remains the canonical source while a draft awaits review.

OpenAI

Official commitmentFulfilledOfficialAI-generated

Astra daily banked-reset commitment fulfilled

On September 3 at 23:12:09 UTC, Tibo promised one banked reset for every day a paid user lacked Astra access. The first grant was initially expected about three hours later. CodexRunway recorded the first delivery on September 4 at 02:30 UTC, without an X completion post. The daily compensation commitment has now been fulfilled.

Expected: 2026-09-04

Official commitmentPartially fulfilledOfficialAI-generated

Tibo says more Codex resets are coming next week

On September 26, 2026 at 14:41 PT, Tibo told a user that more resets were coming next week. Delivery of a new banked reset was reported on September 29, and Pro 200 transition compensation has also arrived. The signal for further resets remains active at 80/100. The original post did not specify all reset dates, counts, eligible plans, or types, so the commitment is not marked as fully fulfilled. Additional signal: OpenAI reported elevated errors across ChatGPT, Codex, and the API on September 29, followed by recovery. The incident provides context only; the status notice does not promise an additional reset or announce its delivery time.

Expected: Date not set

Reported planFulfilledOfficialAI-assisted · reviewed

Codex weekly quota anomaly watch

On October 16–17, 2026, GitHub issue #54677 and multiple independent reports documented sudden weekly Codex quota drops: (1) ~4x used-percentage jump across at least three cases (#54677 rose from 23% to 90% used in 38s with ~381k tokens; #54704 rose from 24% to 96% used with no local model requests and unchanged plan_type=prolite; #54722 rose from 23% to 94%), consistent with a server-side ~4x reduction in effective weekly quota limit without resetting accumulated usage; (2) tight temporal clustering around October 16, 18:09–18:12 UTC across #54677, Reddit Vulcan25 (57%→0% remaining), #54767 (48%→0%), #54704 (76%→4%), #54718 (57%→0%), and Re1ne99 (54%→0%); (3) while prior precedent (#39167) and recent Pro 200→Pro 100 downgrades suggest a batch server-side entitlement/quota update, reports with unchanged plan identifiers (#54704) and post-reset rapid depletion (#54767, #54750) remain unverified pending server-side ledger reconciliation. At the site maintainer's request, the reset signal index is set to 70/100. This is a watch index, not a probability or an OpenAI reset commitment. No official engineering conclusion, fix, or quota reset/compensation has been announced; reset timing is unknown.

Expected: Date not set