Built for people who need evidence, not vibes.
ResetPulse combines fast public-signal collection with an explicit evidence boundary. It is designed for the moment when a post, a dashboard, and a personal account appear to disagree.
A distinct job, not a feature race
Private quota extensions are useful for reading one signed-in account. ResetPulse owns the public evidence layer: a portable record of what was observed, when, with what scope, and how another person can verify it.
Evidence before certainty
A public announcement, an early signal, and a confirmed event are separate states. Ambiguous observations stay qualified instead of becoming a confident-looking answer.
Two clocks, one record
Every event keeps source time and first-observed time separately in UTC. That makes monitoring latency visible instead of hiding it behind a single timestamp.
First-seen history, not just a live hash
The current receipt is a response-level integrity check. When the monitor accepts a source event, it also keeps a write-once first-observed receipt and labels historical backfills, so reviewers can distinguish a live recalculation from the monitor's recorded first record.
Planning without credentials
The forecast is a transparent public-signal baseline. The personal planner and banked-reset reminder stay in the browser and do not ask for Codex credentials.
Action advice without account leakage
The browser-only decision helper can weigh urgency, remaining usage, and banked-reset availability, but it never uploads the inputs or turns a public reset signal into a personal account verdict.
Machine-readable and shareable
The timeline is accompanied by RSS, status, forecast, monitor-health, evidence-receipt, and thresholded community-observation endpoints. Receipts and anonymous outcome summaries can be copied in English, and receipts link to a public schema so people can verify or build on the record.
Forecasts with an audit trail
The public-signal baseline shows its sample size, descriptive backtest accuracy, and Brier score. Lower Brier score is better; a small sample remains a limitation, not a marketing claim.
Privacy as a product feature
Email alerts are optional. Community reports use fixed choices, UTC normalization, rate limiting, and a minimum public sample; no login, account ID, free text, or credential is collected.
Honest coverage labels
A fresh check means the collector ran recently. It does not mean an independent source exists, and official service status is contextual rather than reset proof.
A visible source relationship ledger
The health receipt distinguishes a primary public record, service context, community corroboration, same-origin relays, and accepted independent origins. Two fetches of one origin do not become fake corroboration.
Source health you can inspect
The public monitor-health receipt reports the last successful source fetch, HTTP result, parsed-item count, and fetch duration without revealing the private source URL. A fresh collector and a fresh source are visible as separate facts.
Corroboration without crowd noise
Public Reddit discussions are filtered for explicit global, all-user, or broad usage-limit language. Ambiguous posts about paid reset features are rejected or downgraded to context, so community chatter never silently becomes reset proof.
Claim scope is a first-class field
Every public record now exposes a machine-readable scope claim: broad public, plan-specific, or unknown. The label describes what the source says—not what every private account must have experienced.
Event types stay separate
Full refreshes, banked credits, announcements, and unknown records are classified independently, so the history and forecast do not quietly count an upcoming promise as an executed reset.
Where it fits
Some tools optimize for automatic personal quota reads, desktop badges, or more notification channels. ResetPulse is deliberately different: it makes the public record inspectable, keeps public-signal timing separate from personal quota state, and shows what the monitor does not know. The live timeline also accepts thresholded account-outcome observations, so a public signal can be compared with what users actually saw without pretending that an anecdote is global proof. The live health receipt reports independent corroboration separately, so a high public-evidence score is never presented as private account proof. See the honest comparison before choosing the right layer.
What remains outside the claim
ResetPulse is independent, not an official OpenAI product, and cannot read a private account. A public event may arrive before or after a particular plan or account changes. The data-source notes and methodology are part of the product, not legal fine print.
Try the proof surfaces
Start with the live timeline, open the machine-readable latest evidence receipt, verify it in the browser receipt checker, inspect its public schema, inspect the monitor-health receipt, or subscribe to the RSS event feed. For a practical decision, use the local account decision helper, then keep personal timing in the browser-only planner.