the launch screen

today's screen

The operating point stamped on today's scores: admit rate 8.5%, clean precision 100.00% (95% CI 99.95%–100.00%) — validated 2026-08-26..28 (seeded history, B1).

Mayhem-mode creates sit outside the population the precision numbers were earned on — 3323 of 14168 today (23.5%). They are labeled, never blended in.

dregg_screen scores/2026-08-29.jsonl, UTC day 2026-08-29

today's public sample — most recent CLEAN admits

The holder feed streams every verdict live with /screen lookups on demand; this is the public teaser, refreshed with each publish.

what the verdicts mean

• CLEAN — all five validated gates pass, on chain-exact hydrated features — an absence-of-known-operators verdict, not a prediction of upside
• BUNDLED — two or more birth-slot buyers — the on-chain bundle shape, no Jito id needed; committed operators, fat tails both ways
• NOT_CLEAN — fails remaining gates (e.g. dev buy ≥ 2% of supply)
• KNOWN_CREW — a named fingerprint: crew Jaccard match, recorded rips/dumps, or a recidivist sniper — the most common verdict, not a rare alarm; it names the actor, it does not convict the coin
• UNSCORED — nonstandard curve, hydration failure, or budget policy — reason attached; mayhem-mode launches stay unscored by design

CLEAN is only ever emitted from hydrated, chain-exact features. Scores rank risk; they do not establish intent — the emitted line says what was measured, never who somebody is.

how long the verdicts live — safety and longevity point in opposite directions

A verdict ranks rug risk; it says nothing about upside, and the measured lifetimes run the other way. the modal CLEAN launch fades quietly inside six minutes — median last trade 5.7 min after birth, 16.4% still trading at 6h, collapse by 24h 0.03%, graduation 0.19% (verdict-survival study, CLEAN cohort n=8,773, fresh corpus 2026-08-26..28). CLEAN means nobody with a record is at the table — including anyone who would push it: a safety statement, never a buy signal. Meanwhile BUNDLED coins are the longest-lived, most-graduating cohort AND the most collapse-prone: median last trade 10.3 min, collapse by 24h 3.89% (130× CLEAN), graduation 13.49% (71× CLEAN) (verdict-survival study, BUNDLED cohort n=965, 2026-08-26..28). And KNOWN-CREW is the modal verdict — 85.7% of fresh launches — and those coins mostly just die fast, median last trade 184 seconds (verdict-survival study, 91,505 deployer-identified standard launches, 2026-08-26..28) — the common case, whose card names the actor rather than predicting the coin's path.

mayhem launches are labeled, never scored

every mayhem launch mints 2× supply and parks half in one global pump-operated vault — the same address on all 30,831 coins — whose fee-exempt agent starts selling into the curve a median 2 seconds after birth, re-marking virtual reserves at roughly 500× each trade's real SOL (mayhem-arm study + decoded live tapes, 2026-08-26..29). For those first 24 hours the quoted price is administered, not discovered, so the screen labels the stratum UNSCORED instead of pretending its measured accuracy transfers — the research page carries the full measurement.

why the gates are these gates

coins bundled at birth ripped 1.65% of the time against 0.055% for unbundled — 30× (fresh corpus 2026-08-26..28: 20,895 bundled (≥5 birth-slot buyers) vs 45,322 unbundled). And the repeat offender is not the deployer wearing a new wallet — sniper crews reuse wallets: same-deployer birth-slot sets overlap at mean Jaccard 0.26 against a 0.0026 day-matched null (fresh corpus 2026-08-26..28; degree-preserving null 0.0075). The screen's five gates are the conjunction that survived validation, not a score we liked. Two weeks later the ledger still knows them: crew fingerprints are durable: the 400 busiest returning deployers' new launches matched their own recorded crew 48.5% of the time (strangers: 0.59%), and fingerprint overlap retains 93% of its same-day strength after two weeks (crew-persistence study, window A 2026-08-05..14 to window B 2026-08-26..28, 11 unobserved days between).