# wlink live-readiness note for Ben (exec, 12 Oct)
Verdict: auditor2 VERIFIED small size (http://159.223.88.168/hx-c9ois/fw/audit-wlink-verdict.md). Not live yet; nothing switched.

**Rule (arm A):** Uniswap v3 fee-100 WETH pool, pool cap < $150k. When the 3rd distinct "winner wallet" (early buyer of an earlier runner, 6 h lag, infra wallets excluded) buys within 180 s of pool start, buy at the first print >= 10 s later. Exit: take-profit 5x, stop 0.5x, time stop 4 h.
**Evidence:** replay n133 pull-adjusted mean 1.81x (80% band 1.58-2.03); forward n36 mean 1.80x. Median is below 1 (0.82-0.92): a lottery profile, about a third of trades hit 5x, ~30% of pools get pulled (counted as 0x).

**Size:** start at the current $10 per trade, max 2 wlink positions open at once. Paper mean 1.8x is about +$8 per $10 trade, but only ~1/3 of trades win, so expect long red streaks. Do not raise size until 30 real trades.

**Fail-safes already in fmmc5 (v3 trader):** buy+sell simulated by eth_call through the real pool before every buy (skip if it fails or round trip < min_roundtrip); stop 0.5x with emergency slippage; kill switch (kill_dd_usd 100, after n=20 kill if mean < 0.5); Ben-only LIVE file and size.

**Gaps before live (exec code work, about half a day):**
1. fmmc5 enters by market-cap window (250k-1M), not by wallet signal. wlink needs a second entry lane fed by the test manager's signal (pool, signal time), pushed to firstmin as a file, not queried live.
2. Thin pools (20-60k): pre-buy round-trip sim must be strict (min_roundtrip 0.8); replay assumed 3% fee + 5% slip.
3. Funder coverage only 39%: a single operator could still bias the signal. Watch the first 20 real trades against the replay band; stop if mean < 0.8x.

**Ask:** approve building the wlink lane in paper mode on firstmin first (no money, 2-3 days), then bring Ben the live switch with fill-vs-quote numbers.