# A1 iterate run 18 (10 Oct 2026): clocks and time exits on the new native tape

Data: /opt/ppcopy/r6nat.db (native-ETH v4 pools, 21 Sep to 9 Oct 21:30 UTC, built by ppdata), A2 loader /tmp/a2_nat.pkl, my WETH P2.pkl. Sim: pull-aware (liq <= 10% of max = 0), 3% slip each side, pool fee, exit needs in-range ETH >= 2x position, position 0.1 ETH. Entry impact checked: concentrated LP is deep, 0.1 ETH moves price 2-3% only. Scripts ba1-ba30.py in /opt/crew/work/analyst1.

## Finding 1: native H3000 (fee 3000, hook) runs on a CLOCK, and the hook code tells which clock
- Hook runtime code size/hash (eth_getCode, 330 hooks): 16,724 B (hash 635e21...) = 116 per-pool hooks, 2-8 Oct. 17,916 B (8ea6a1...) = 56 hooks, 21-30 Sep. 21,824 B (7f3304..., ONE shared hook 0xfcd182) = 67 pools on 9 Oct, dead (no ramp).
- Version 16,724 B (n=115): all pools have exactly ONE recipient, ~497 swaps in the first 600 s, ETH reserve 16.7-19 ETH at 600 s. One actor ramps the price; no outside buyer. Price peaks at pool age 640 s (IQR 633-680) and the operator pulls at ~1,325 s (52% of pulled pools in the 1290-1340 s bin).
- Old version 17,916 B: peak at 1,167 s (IQR 750-1258). So the timeline is per code version.
- Exit by clock beats the TP ladder (unique-hook native, entry first swap +5 s, n=245):

| exit | mean | halves | zero |
|---|---|---|---|
| tp3 / 4 h | 2.17 | 1.83 / 2.51 | 21% |
| tp5 / 4 h | 3.23 | 2.36 / 4.09 | 29% |
| tp20 / 4 h | 1.65 | 2.17 / 1.13 | 82% |
| tp50 / 4 h | 1.37 | 2.21 / 0.54 | 85% |
| sell at age 600 s, no TP | 5.20 | 3.93 / 6.46 | 22% |
| version 16,724 B only (n=115), sell at 600 s | 7.28 | | 1% |

  Age-600 curve (entry +5 s): 300 s 2.74, 450 s 3.16, 540 s 4.36, 600 s 4.94, 630 s 4.97, 660 s 4.81, 720 s 4.17, 900 s 3.65, 1200 s 3.50, 1300 s 3.16 (zero 38%).
- Entry speed is the cliff: price is ~1.00-1.03 for 5 s after the first swap (dust), then ONE block of swaps moves it to 1.75, 2.3 at +10 s, 3.1 at +60 s. Version 16,724 B sell at 605 s: entry first swap +0..3 s 8.7-9.1x, +5 s 7.3x, +8 s 5.1x, +10 s 3.9x, +30 s 3.7x. Creation to first swap is 5.4 s (p10-p90 4.6-9.2). Landing before the burst needs the buy inside ~10 s of pool creation; after the burst it is still 3.7x.
- Decay: age-600 mean by day for unique hooks: 4 Oct 7.4, 5 Oct 6.9, 6 Oct 6.1, 7 Oct 4.1 (zero 44%), 8 Oct 2.9. On 9 Oct (v4tape since 06:40 UTC) 8 pools of code 635e21 and 67 of the dead shared hook. Of the 8 live ones 4 followed the ramp (6.0-9.5x at 600 s, pulled ~1,262 s after first swap), 4 did not (0.4-0.5x or flat). Lifecycle risk: the operator is switching template.
- Risk, unchanged: nobody else trades these pools, so whether an outside wallet can SELL is only testable by the real sell test (A2/Ben). Time exit makes the proof cheap: a $5 buy and a $5 sell 10 minutes later.

## Finding 2: time exit beats TP on the other fee-100 families too
Entry +5 s, tp off, sell at fixed age (pull = 0):
- WETH u100 (n=570): sell at 120 s 1.45 (halves 1.37/1.52, zero 38%), 300 s 1.50 (zero 54%) vs tp3/4 h 1.31 (zero 55%).
- Native fee100 unhooked (n=1271): sell at 120 s 1.29-1.30 (halves 1.26/1.32), 60 s 1.19, 300 s 1.14, vs tp3 1.09 (A2). Days 0.9-1.8, 14 of 19 days above 1.0 (small days are noisy). Zero 41%. Median 1.18.
- WETH h100: TP better (tp3 2.43 vs 120 s 1.46, 300 s 1.88). Keep tp3-4.
- WETH H3000 (v43000): no clock (pulled after 10,000 s in 118/135); hold 1 h no TP 22x, 4 h 41x (known family, A2 ladder).
- Native fee3000 unhooked: 1.62 at 600 s, zero 52%: weak. Native fee100 hooked: 87% zero, pulled at median 22 s (29% right when the next pool starts): never.

## Discards this run
- Code-hash lineage prior (tokcode2, expanding mean): equals the family label (h100 386 of 386 prior >= 1.5; u100 prior < 1.5 1.01), covers 576 of 3097 pools.
- Day-2 re-entry of coins that crossed $500k (runtrack, n=1053): +1 h 0.98, +24 h 0.93, median 1.00 (stale caps). No drift edge.
- Native fee100 pull inversion at +5 s: AUC reserve 0.41, swaps 0.63 (halves 0.67/0.56), rest 0.5. Not separable.
- Peak-time clocks for h100, u100, v3100, v43000, v4dyn, native unhooked: no day with a tight mode (only native H3000 and nat100hk day 20733).