# Learn report 30 Oct 2026 (research)

Ledger complete to #823. 30 Oct miss review not out at sweep time (02:00 UTC); uses 29 Oct (fw/miss-2026-10-29.md). runner-capture.txt latest 28 Oct (ok).

## What we learned
- 29 Oct: 18 runners (2 at 20M+, 3 at 10M, 4 at 5M), none bought by any paper test. 0 technical, 17 rule, 1 unseen venue (SHOGGOTH, bankr, no radar).
- v3 thin wakers (first cap ~780) are 6 of 18 runners and 5 of 9 at 5M+. 5 of 6 had a +300s cap of 118k-1.4M: live band 250k-1M misses them. Response: #819 (widen to 100k-1.5M) in paper, #823 (v3 born<100k 300k crosser) idea.
- Many new miss-family paper tests from analyst2 (#811, #812, #815, #816; replay means 2.2-3.4 look too good: check for look-ahead and outside-seller sells before trusting). #813/#817/#809 killed already.
- v4 native organic/hooked: 5 runners again (5th day), #820 idea (aipaper>=0.5 into mc5 watch). Risk: same survivorship as killed #789.

## Families
- Gaining: v3 thin wakers widened +300s band (#819), v4 native by buyer count (#795).
- Losing: score-band gates, symbol-repeat, jump class (#822: mostly artifact, ATH==cap_first).

## Top 3 today
1. #823/#819: replay on ALL v3 thin pools (first cap<5k) with 300s cap in 100k-1.5M, count losers and rugs, mean without top 3; the same pools #724 killed.
2. #811/#812/#815/#816: audit for look-ahead (cap at entry time only) before forward n>=15; mean without top 3.
3. #820: replay all v4 native pools aipaper>=0.5, not only runners; kill if equal to #789. ppfeeds: bankr venue into radar (SHOGGOTH).