# Learn report 14 Oct 2026 (research)

Ledger ~579 rows. Newest (#574-#579) are miss-family ideas, all with rows. #576 (v3 bundle bb>=3) killed: artifact of launcher 0x73991a, no edge outside it. Seedhi rows #493/#537 are killed (closes the old open item).

## What we learned
- **Live money still flat** (fw/realpnl.txt, 14 Oct): 63 trades, 0.98x per ETH, net -0.0072 ETH. $10 lane +0.159 ETH (50), $50 lane -0.166 ETH (13). 12 near-total losses cost 0.079 ETH. Days: 11 Oct -0.012, 12 Oct -0.036. Shadow last 3 days n=19 mean 1.82x, up from 0.43x yesterday (small n, do not trust yet).
- **Misses 13 Oct** (fw/miss-2026-10-13.md): 27 runners missed, none unwatched. 10 of 27 are one native fee100 nohook seed-25k kit (cap_first 24-25k, ATH 1-2M); mc5 does not watch that venue. 19 are 1M runners, 4 at 10M+. v3 born-big pools (casino, CAPITOL, VERDEX, BONER) hit 1.4-3.6M cap by minute 5, past mc5's rule.
- **Runner capture** (runner-capture.txt): 12 Oct 8 runners, live bought 4, any bot 8; peak kept still 0%. Catching is now better than keeping.
- **Exit leak unchanged** (runner-exits.md): 26 of 71 live trades peaked >=3x; most of the big ones (35x, 35x, 29x, 24x, 22x) ended at the 0.5x stop (1% of peak kept). The no-stop gain is mostly 3 giants, not yet proven once rugs are priced in.
- **Data**: review/misses.csv is fresh again (00:41 today); the 3-day gap is closed. a3: the first-block bundle marker is real as a runner marker but not tradable with current data (no depth, no sellability).

## Families
- Gaining: v3-mc5/exits (only plausible +ETH lever); miss/native-kit (blocked on tape, not on ideas).
- Losing: unhooked-v4/blind-spot (83 killed vs 0 alive), v3-mc5/first-minute-filters (21 killed), v3-mc5/wallets (24 killed). Stop spending runs there; analyst "idle run" rows (#573) are noise.

## Top 3 directions today
1. Native fee100 nohook kit (#561/#577/#579): the single biggest miss cluster. Blocked on v4tape keeping A1 pools 90 min (asked ppfeeds). Once 3 days of tape exist: replay with outside-wallet sell check, loser leg sized, holdout 12-14 Oct. Do not trust the 49% hit rate until the loser leg exists (decoy-kit risk).
2. mc5 exit with rug cost: one audited rule (delayed/wider stop) on all mc5 buys, report mean, without-top-3 and rug loss. Still the best use of live-money effort.
3. $50 vs $10 lane: explain the difference (entry cap, day, rug) before any size review.