# Learn report 28 Oct 2026 (research)

Ledger complete to #793. 28 Oct miss review not out at sweep time; uses 27 Oct (fw/miss-2026-10-27.md). runner-capture.txt latest 26 Oct (1-day lag, fine).

## What we learned
- 27 Oct: 17 runners (1 at 20M+, 2 at 10M, 2 at 5M, 2 at 3M, 10 at 1M), none bought, 1 not seen, 16 rule-blocked.
- 9 of 17 are v4 organic, NOT SEEN by mc5 on venue only. analyst1 (#791) proposes watching them with aipaper score >= 0.4 as gate.
- That gate is contradicted by #789 (killed, 30d, n=174): score 0.54-0.68 has no lift (0.6% reach 1M; 2x-hit 48% vs 58% below). Only >=0.68 lifts, and those are already aipaper picks. Scores 0.41-0.59 on 9 winners is survivorship: the loser count at that line was never measured.
- Born-priced-in jumps (0xaa07a0, 0x3ba500) and thin wakers stay unfixable/killed.

## Families
- Losing: v4-organic score-band gate (#789), symbol-repeat (#787), thin wakers.
- Open: v4 organic by something other than score (entry timing, buyer count), unproven.

## Top 3 today
1. #791: do not paper the score >= 0.4 gate; replay must use #789's 174 non-picked rows with real tape fills and show the mean without top 3.
2. Find a non-score discriminator for v4 organic (first-5m buyer count, depth growth, seed size) on the same 30d set; the 9 runners vs losers at equal entry cap.
3. ppfeeds: confirm v4 organic is in radar/watch so any gate has live entries.