Reports
A graduation event is not a graduation
Published 11 September 2026. The figures below were computed from the record on that date and are not updated afterwards: they are read from a file written when this report was published, not recomputed when this page is built. The live view is here.
A bonding curve graduating is an event on a feed before it is anything else. We record that event, and then, separately, we go and read the curve account on chain. Across 12,524 graduation events in this archive the two agree for some and disagree for a great many, and which group a launch falls into is knowable at the time rather than in hindsight.
Where a PumpSwap pool exists, the event holds up. Of the 6,092 events backed by a pool, we have re-read the curve for 1,774, and 31 (1.7%) read incomplete. Two independent pieces of evidence agreeing is what a confirmed graduation looks like.
Where no pool exists, mostly nothing happened. 5,489 events have no pool behind them. We have read the curve for 5,378 of them: 5,187 (96.4%) had not completed, and 0 had. A threshold crossed on a feed, and no curve behind it.
What the record held
as published, 2026-09-11| Confirmed by | Feed events | Read on chain | Read complete | Read incomplete | Account gone |
|---|---|---|---|---|---|
| pool | 6,092 | 1,774 | 1,743 | 31 | 0 |
| (none) | 5,489 | 5,378 | 0 | 5,187 | 191 |
| curve_complete | 943 | 943 | 943 | 0 | 0 |
The row you must not use
The middle group is circular and is printed anyway, because leaving it out would be the more misleading choice. graduated_confirmed_by = 'curve_complete' means the curve read is the confirmation: those 943 rows are complete by construction and can appear in no rate about curve readings without making it say what it was built from. The comparison above is between the pool-confirmed group and the unconfirmed one, because a pool existing is evidence the curve read had no part in producing.
What this does not say
It does not say these tokens are frauds. A curve that has not completed is a curve that has not completed, and launches fail at that stage constantly and innocently. What it says is narrower and more useful: a graduation event, taken alone, is not evidence that a curve finished, and roughly two in five of them in this archive have nothing else behind them.
It is also not a rate over graduations in general. We read the curve for unconfirmed events far more often than for confirmed ones, deliberately, which is why the two populations are reported separately and never pooled into one figure.
A curve can complete long after launch, so a single incomplete reading does not settle a launch forever; the readings are retaken on a cooldown and the published columns are rewritten from the collector on every build. The counts here are what those columns held on 2026-09-11.
Coverage began 2026-09-02 12:29 UTC and the figures were taken from the record built 2026-09-11 17:12 UTC. Launches before then were not watched. Rows a detector restored after the fact, and rows rebuilt from chain history, are excluded throughout: a launch found late shows no outside buyers because nobody was watching it, which would flatter every figure here.
Check it yourself
One query against the record database. Run it today and you will get a larger answer than the table above, because the archive has grown since publication, and that is the difference between a report and a live view, and it is why both exist. Disagreeing with either is the point of publishing them.
| SELECT COALESCE(graduated_confirmed_by, '(none)') AS confirmed_by, COUNT(*) AS feed_events, SUM(curve_checked_at IS NOT NULL) AS read_on_chain, SUM(curve_complete = 1) AS read_complete, SUM(curve_complete = 0) AS read_incomplete, SUM(curve_checked_at IS NOT NULL AND curve_complete IS NULL) AS account_gone FROM tokens WHERE graduated = 1 GROUP BY confirmed_by ORDER BY feed_events DESC; | the table above, verbatim, as it stood on 2026-09-11. Bulk file: record.db, with an API key |
Since 13 September 2026 the record database is licensed and downloads with an API key. Text above written before that date is left as published.