--- name: trading-edge-hub-10 description: Trading edge hub ledger part 10 (continues trading-edge-hub-9, near its cap): from 6 Oct 2026 17:45 BKK. Read with [[trading-edge-hub]] and [[trading-edge-hub-9]] before any buy/sell call. sources: [cowork] --- - 2026-10-06 17:43 BKK: A2.24 honest money v2 with LIVE kit-hold exit (wide gate, gate+30 s, 0.3x stop, block-end exits, real depth, slot rule). $1k: +$3,317 / +$6,683 a day (vs +$2,733 / +$3,535 all-out at 5x), worst day and max DD unchanged (-$4,257 / -$4,700). With the 2 still-open holds valued at 5x sale, half B +$5,446. $500 +1,745 / +3,511; $2k +6,002 / +12,138; $3k +8,153 / +16,577. 15 kit holds: Kit A 9, Kit B 6; 4 of 15 ended under the 5x sale (worst 0xb91b4020518cd0cf824dca8291e94cd62c7f0b91 at 0.65x). Peak open 5 tickets. Wallet 8x ticket + 0.005 ETH enough on history (max DD 5.7 tickets); 15x needed to guarantee the 10-ticket kill switch always trips before cash runs out. - 2026-10-06 17:47 BKK: KILLED Virality V10 (buy at hour 8 after $500k on slow-burn signal): $1k -$519 / -$250 a day, 0.40x / 0.42x, no trade reached 4x after hour 8. V4 "slow burn predicts $20M+" RETRACTED: 11 of 50 coins had overflow max caps in pp3, 15 of 34 slow-burn coins were already $2M to $40M at the $500k crossing. V3: ~80% of coins are "flash" (buyer wave done within hours); the spread happens before or around the $1M gate, so the live gate already sits where the money is. Lesson: use max cap / cap at crossing, never "reached $X", on pp3. - 2026-10-06 17:50 BKK: Ben said start live at $100 a trade to test. Coder v1.33: hard ticket cap $100 until Supervisor writes LIFT CAP. Funding: 0.30 ETH to 0x04a057D15bd07581DC88b0a37E41EB69Ceb8C2D3. - 2026-10-06 17:53 BKK (FW Supervisor) | fmmc5 LIVE feed hole 17:39 to 17:40 (three 21.6 s gaps, under the old 30 s stall rule, ~1 min blind vs 6 s fill pass line). Root cause (PP): btime tried public RPC first with a 20 s timeout. Fix md5 8a806081ee9ec84c07fc26248268cf90: btime via drpc-first feed_res, 4 s call timeout, FEED_FLIP on any round >5 s or 2 gaps >8 s in 2 min (5 min on next source). 0 open live trades at restart. A2 v20 re-audits diff + 15 min proof | TRAP for every live bot: a failover threshold above the fill pass line never fires when it matters | OPEN - 2026-10-06 17:57 BKK (FW Analyst 1 A1.24, sup-to-a1 v20) | mc5 TOP CELLS FORWARD PAPER live: http://159.223.88.168/hx-c9ois/fw/a1-mc5-top-paper.txt (research-box cron :05/:35 mc5top.py, main pulls */5): +5 min $1M-$2M, +10 min $500k-$2M, +10 min $1M-$3M from 6 Oct 17:46 BKK, dust guard, 6 s fill, 0.5x stop, 4 h, base wins; pass check 8 Oct 17:46 BKK (positive AND positive without top 2) | AUTO LIVE MATCH: http://159.223.88.168/hx-c9ois/fw/a1-live-match.txt (livematch.py */5, replays each closed fmmc5 coin, MATCH within 0.05x) | TRAP: fmmc5 watch prices a stop on the crossing swap (0x4fdcc31f5514644bf512b296057f025ddd4f9308 0.52x) vs replay 3 s later 0.42x and v3mc5 0.41x; real live sells land later, expect the lower number | LIVE (paper) - 2026-10-06 17:57 BKK (FW Supervisor) | A1 v21 (sup-to-a1 md5 67bfdd7fe98d79e54db43b3d863bb562): DEAD COIN EXIT test on mc5 (79% lose, mean loser 0.63 ticket, 19 of 116 stops fill under 0.3x): at +10/+20/+30/+60 min after fill sell if price < 0.7/0.85/1.0 of entry AND new buyer wallets last 10 min < 0/3/10; pick H1, test H2; pass = H2 $/day up, worst 3 days no worse, up without top 2 | OPEN - 2026-10-06 18:06 BKK (FW Analyst 1 A1.25, http://159.223.88.168/hx-c9ois/fw/a1-mc5-dead.md) | mc5 DEAD COIN EXIT (178 dust-guarded trades, 6 s fill, every exit at end of next block, 48-cell grid T +10/20/30/60 min x price <0.7/0.85/1.0 of fill x new buyer wallets <=0/3/10 or none): H1 pick +10 min price 0.51 ticket) but kills 6 winners incl. 0x6e8a6a1af8ec53f3f09c7303e5403558dee42d0a 35.88x and 0xcdc10abe9585ce4127ac4f4c1fe17daf34b0be7a 15.20x | buyer test never fires (median 472 new wallets first hour, dead coins too) | cheap losers = pulls and one-block dumps, no slow signal | KILLED; mean loser 0.63 ticket stays the wallet number | FACT (paper) - 2026-10-06 18:08 BKK (FW Supervisor) | A1.25 dead coin exit KILLED (accepted). A1 v22 (sup-to-a1 md5 f304333c126bad9a42aa196ac8ba812e): LP OWNER check at +5 min (pull needs the LP owner): class deployer/fresh EOA vs locker/contract vs burned, plus top owner share of in-range units; per-class pull rate and $/day; test skip worst class and 1.5x/0.5x size tilt; also NFT transfer/approve before Burn as early warning | OPEN - 2026-10-06 18:07 BKK, server clock (FW Analyst 2 v20 status md5 744d8d6a6b9b9c0c56fac87f9c86581d) | fmmc5 8a806081ee9ec84c07fc26248268cf90 STILL READY YES: diff vs fe43786dd912c04d426b7e8994832d3e touches only btime/feed_res/feed_flip/round timing (entry, exit, kill, gas, live buy/sell unchanged); 17:50 to 18:07 max gap 5.9 s (was 21.6 s), 0 flips, 0 stalls, 1 failed call (dRPC timeout at 18:02 fell back to public in the same round); dRPC calls +65% (btime) | BUG (old, kill path): every restart re-runs LIVE_ON and resets start_t, so mc5_dd forgets past live trades and the 20-ticket kill restarts at 0; fix = start_t from LIVE_MC5 file mtime; must ship before $100 step | TRAP for every live bot: a drawdown kill anchored to process start is wiped by every deploy | OPEN - 2026-10-06 18:10 BKK (FW Supervisor) | fmmc5 feed fix CLOSED (server md5 8a806081ee9ec84c07fc26248268cf90, 18:08 max gap 2.1 s, 0 flips; LIVE_MC5 on, 0 live trades). PP v26 (sup-to-pp md5 2393f46c79cf9edfc0a26a9afa4dae94): fix A2's restart kill bug: anchor live session to LIVE_MC5 mtime / persisted live_start so the 20-ticket drawdown survives restarts; prove dd unchanged across a test restart; must ship before any size step | OPEN - 2026-10-06 18:13 BKK (FW Supervisor) | TRAP (publish route): a hand-copied gzip+base64 patch can corrupt silently: gunzip CRC error, then patch applied a garbled half text (sup-to-pp v26 went out unreadable, PP rung). Fix: check md5 of the b64 on the Mac before applying and abort on mismatch; for short edits send PLAIN TEXT via ssh stdin + python line replace, then verify final md5. sup-to-pp v26 repaired to 2393f46c79cf9edfc0a26a9afa4dae94, PP re-rung | FACT - 2026-10-06 18:12 BKK (PP, answer to sup-to-pp v26) | fmmc5 restart kill bug FIXED, md5 b4154a937762a6929abc146aa6fc8cec (restart 18:10:59, 0 open live, self test PASS): live_session() anchors start_t to LIVE_MC5 mtime (kv live_file_mt); same file after restart = LIVE_RESUME (start_t + start wallet kept), new/recreated file = LIVE_ON new session. Copy test PASS (5.0 tickets before and after restart, old code 0.0; recreated file 0.0), http://159.223.88.168/hx-c9ois/fw/pp-mc5-session-test.txt. Live board now "since LIVE_ON 06 Oct 17:26:37" (17:49 restart had moved it to 17:49:57) | RULE for every live bot: persist session anchors outside the process, keyed to the operator's on-switch | CLOSED pending A2 - 2026-10-06 18:13 BKK, server clock (FW Analyst 2 v21 status md5 13c4e80c5d16f236dc653c686321c730) | fmmc5 b4154a937762a6929abc146aa6fc8cec STILL READY YES: diff vs 8a806081ee9ec84c07fc26248268cf90 = only live_session() + its call in main(); entry, exit, feed, gas, live buy/sell untouched; 18:11:00 LIVE_RESUME, board drawdown "since LIVE_ON 06 Oct 17:26:37" (LIVE_MC5 mtime 1791282397); feed 18:12-18:13 max gap 3.1 s, 0 failed | restart kill bug CLOSED | TRAP: a `touch` of /opt/fm/LIVE_MC5 starts a new session and resets drawdown; only create or delete it | FACT - 2026-10-06 18:17 BKK (SW Analyst 2 A2.25, http://159.223.88.168/hx-c9ois/sw/sw-a2-25-more-trades.md) | MORE TRADES via gate knobs (live kit-hold exit, slot rule, gate+30 s, next-block exits, real depth, $1k, rule pre-registered): only sellable $750k passes (+$3,973 / +$7,177 a day vs live +$3,317 / +$6,683; worst day -$3,713 vs floor -$3,951) but half B max DD -$4,700 -> -$7,041. The 22 extra coins LOSE (-$280 / -$518 a day); all the gain is the SAME coins bought earlier, only +$302 / +$113 a day without the top 2 per half. $500k, age 6/12 h, top 0.10/0.12, army 0.02/0.03 all FAIL; no pair beats $750k alone | live gate already sits where the money is: loosening adds only losers | KEEP LIVE GATE; $750k paper variant only | FACT (paper) - 2026-10-06 18:18 BKK (FW Analyst 1 A1.26, http://159.223.88.168/hx-c9ois/fw/a1-mc5-lpowner.md) | mc5 LP OWNER at +5 min (178 trades): largest in-range holder is ONE plain wallet (EOA) in 162, holds 100% in 176 | EOA class carries the money (H2 2.41x, +$1,599/day at $100); LP in contract 8 trades 0.60x; LP burned 8 trades (H2 only) 0.43x, never pulled, still lose | SKIP CONTRACT (H1 pick) PASS but small: H2 +$1,561 vs +$1,550/day at $100 (+$12,832 vs +$12,658 at $1k), worst 3 days equal, no-top-2 up, 3 H2 trades | size tilt to locked/burned FAILS (+$1,263) | NO NFT early warning: 1 of 85 pulled pools had an approve before the Burn (0xaf36a2b5f81ab17c194462654e211b569a1f3cdc, 1,834 s), 28 of 93 not-pulled had moves | mechanism: the dev who can pull is the dev who pumps; locked/burned LP coins do not run | CANDIDATE (small filter), FACT - 6 Oct 18:25 BKK, SW A2.25 (KILLED, gate knobs): loosening the live second-window gate one knob at a time (sellable $750k/$500k, age 6h/12h, top receiver 0.10/0.12, army 0.02/0.03) never adds paying coins. Only $750k passed (+$656/+$494 a day at $1k), and all of that came from buying the SAME 67 coins earlier; the 22 extra coins lost (-$280/-$518 a day). Gate stays $1M/3h/0.08/0.015. Open: A2.26 tests whether the $750k entry can be split using only what is known at $750k (kit tag first). Coder v1.35 adds a paper book w2_v3_750k. - 2026-10-06 18:22 BKK (FW Supervisor) | A1.26 decision: NO live change (skip-contract +$11/day on 3 H2 trades = noise); A1 tags lp_class in live match + paper logs, re-test skip contract/burned at 15+ trades each. A1 v23 (sup-to-a1 md5 73dc1024e44bbb6d719ef8b6b197347c): CONDITIONAL HOLD past 4 h only if x >= 2/3/5 AND new buyers last 60 min >= 50/150/300 AND price above 1 h ago, hold to 8/12 h (stop 0.5x of 4 h price) | OPEN - 2026-10-06 18:22 BKK (SW Analyst 2 A2.26, http://159.223.88.168/hx-c9ois/sw/sw-a2-26-early-entry.md) | EARLIER ENTRY WITHOUT LOSERS: at the $750k touch, buy only if a split says GO, else wait for live $1M (live exit, slot rule, all frictions, $1k, pre-registered): 0 of 25 tries pass (kit rule + 11 features at median both ways: buys/buyers 5 and 15 min, net buy $, launch to $750k, top, army, depth, fresh-wallet share). 13 of 25 beat live in both halves = 52%; RANDOM half picks (200 draws) do it 52% of the time; both sides of the cut beat live for buyers15 and depth = noise. Kit A or B at $750k +$3,601 / +$7,137 vs live +$3,317 / +$6,683, plain $750k +$3,973 / +$7,177 | METHOD: always test a split against random picks of the same size; if both sides of a cut win, it is not a signal | CLOSED, keep live $1M | FACT (paper) - 6 Oct 18:30 BKK, SW A2.26 (KILLED, early entry split): no feature known at the $750k touch (kit tag, buys/buyers 5 and 15 min, net buy, time to $750k, top receiver, army, depth, fresh-wallet share) beats a random-half pick: 13 of 25 tries beat live, random does 52%. Method to reuse everywhere: random-half baseline (200 draws) before trusting any filter. Next: A2.27 non-kit exit grid; Virality V12 reruns the hour-8 slow-burn entry on real buyer data (pp3-series with buyers, md5 9972baf44632b553861149f50ef483c8). - 2026-10-06 18:27 BKK (SW Analyst 2 A2.27, http://159.223.88.168/hx-c9ois/sw/sw-a2-27-nonkit-exit.md) | NON-KIT EXIT (34 non-kit coins, kit hold unchanged, $1k, all frictions, slot rule, pre-registered incl. random-half baseline + gain without top 2): every target change loses vs live +$3,317 / +$6,683 a day (3x -$947/-$611, 4x -$761/-$301, 7x -$109/+$321, 10x -$1,276/+$913, 80% at 5x + 20% runner -$213/+$48, half at 3x -$423/-$285); stop 0.25x -$38/-$12. Stop 0.4x passes by a hair (+$81/+$27, only better stop prices) but REJECTED: kit winners 0x9ba4d84ac1571a0788b2bb7a6412e73529d12dc6 dipped to 0.308x and 0xe99c8437b0b29325a4080b8c837ff4e5a6f9fc69 to 0.357x before 5x (~+$12k at $1k) vs $1,380 saved | TRAP: a tighter stop "losing no winners" on a small sample is luck; check the dip-before-target of ALL winners in the book | KEEP LIVE EXIT | FACT (paper) - 2026-10-06 18:26 BKK (FW Analyst 1 A1.27, http://159.223.88.168/hx-c9ois/fw/a1-mc5-extend.md) | mc5 CONDITIONAL HOLD PAST 4 H (open at 4 h, x>=2/3/5, new buyer wallets last 60 min >=50/150/300, price above 1 h ago -> hold 8/12 h, stop 0.5x of 4 h value): FAIL all 18 cells both halves; H1 pick (8 h, M2, B300) H2 $1,212 vs $1,550/day at $100 ($9,560 vs $12,658 at $1k), worst 3 days -$913 vs +$715 | 10 of 11 extended coins gave money back (0xaffa3ccfee7eea8e8fe413c8aecc444acdaa885d 14.60x -> pulled), only 0x6e8a6a1af8ec53f3f09c7303e5403558dee42d0a +$108 | hundreds of new wallets an hour arrive on bleeding coins too = not a hold signal | KILLED; 4 h stays | lp_class tag (EOA/CONTRACT/BURNED) live on fw/a1-live-match.txt and fw/a1-mc5-top-paper.txt via lpclass.py | FACT (paper) - 2026-10-06 18:26 BKK (FW Analyst 1) | TRAP (publish route, confirmed again): hand-relayed base64 of a tarball came out 48 bytes longer on the Mac (md5 check caught it, nothing applied). For text docs send plain text via quoted heredoc and verify md5 on the server | FACT - 2026-10-06 18:30 BKK (FW Supervisor) | mc5 rule now tested on every knob: entry checkpoint, band low and high, stop level, time stop, dead-coin exit, conditional hold, kit split, LP owner, re-entry, quality filters: base +5 min / $250k-$1M / dust guard / 0.5x / 4 h wins every time | RESEARCH SATURATED on mc5; remaining lever = SIZE (money test says edge holds to ~$2k-$3k a ticket) and live fill proof | A1 on standby (live match + paper logs running), next job after the first live trades | FACT - 6 Oct 18:35 BKK, SW A2.27 (KILLED, non-kit exit): for non-kit coins, sell all at 5x is the top. 3x/4x/7x/10x, 80/20 runner and half-at-3x all lose vs live in at least one half. Stop 0.4x passes by a hair (+$81/+$27 a day) but is rejected: Kit B 0x9ba4d84ac1571a0788b2bb7a6412e73529d12dc6 dipped to 0.308x and Kit A 0xe99c8437b0b29325a4080b8c837ff4e5a6f9fc69 to 0.357x before 5x (~+$12k at $1k). Live exit stays 0.3x / 5x / kit hold. Entry and exit both closed; next lever is size (A2.28). Virality V10 slow-burn stays dead on real buyer data (best trade 1.04x). - 2026-10-06 18:31 BKK (SW Analyst 2 A2.28, http://159.223.88.168/hx-c9ois/sw/sw-a2-28-size-v2.md) | SIZE v2 with live exit (kit hold, slot rule, all frictions): $ a day $500 +1,745 / +3,511; $1k +3,317 / +6,683 (smallest for $2k a day both halves); $2k +6,002 / +12,138; $3k +8,153 / +16,577; $5k +11,156 / +23,039; all sizes count (worst day far above -10x ticket). $ per $ wallet highest at smallest ticket, falls every step (no hump); max DD 4.2 to 4.7 tickets at every size; 5x ticket wallet never skipped a trade. Own buy moves pool >10%: 0% at $2k, 89%/79% at $3k, 100% at $5k (gates all at $1M sellable = near-identical depth); x lost to pool vs paper 0.19/0.30 at $1k, 0.61/0.89 at $5k | RAMP $1k -> $2k after live fills match -> $3k top | FACT (paper) - 6 Oct 18:35 BKK, SW A2.28 (size curve with live exit): $ a day $500 +1,745/+3,511; $1k +3,317/+6,683; $2k +6,002/+12,138; $3k +8,153/+16,577; $5k +11,156/+23,039 (half A/B). $1k is the smallest ticket clearing $2,000 a day in both halves. Max drawdown ~4.2 to 4.7 tickets at every size. Own buy moves the pool >10% in 0% of trades at $2k but 79-89% at $3k (every gate sits at ~$1M, so depth is near constant). Ramp plan: $100 test, then $1k, $2k after live fills match paper, $3k top. - 2026-10-06 18:46 BKK (SW Analyst 1 A1.20, http://159.223.88.168/hx-c9ois/sw/a1-20-kit-c.md) | FIND KIT C: all 7,108 v3 WETH pools since 13 Sep grouped by launcher code hash and token code hash, 24 h swaps/mints/burns, base = first swap +5 min (dust guard), peak after first 5x before pull | NO Kit C: no launcher with 20+ launches besides Kit A/B has 10+ 5x coins in both halves | yardstick: peak after 5x median/top 10% Kit A 2.32/7.18 (A), 2.50/9.31 (B); Kit B 1.89/10.40 (B); non-kit 1.61/6.78 (A), 1.74/3.34 (B) -> kit edge is the TAIL (median gap only 1.44x Kit A, 1.09x Kit B) | WATCH: 3,587-byte launcher family (hashes 51d0e39db2ed, 2696b754c8e3, 60341fa374bf), 17 launches 29 Sep to 3 Oct BKK, 10 reached $1M gate (59%), 6 wide trades mean 2.61x, quiet since 3 Oct -> tag kit_3587 | PULL TEMPLATES never reach gate: launcher 0x9b226f198ef0038d054be6265aa9f76c0d6e5e78 (63, 100% pulled 2 h), PM token templates 0402ab611457 (360, 100%), 5fcfacb98bd0 (151, 100%), 27a0e402b29a (76, 95%), 0c17e487a6b2 (201, 77%) | shared launcher 0x22a2097014cd330c27607fb7e3f97901913e0a3e: 53 launches 23 to 25 Sep, 5 gates, dead | KILLED (Kit C), WATCH (kit_3587) | FACT - 6 Oct 18:55 BKK, SW A1.20 (no Kit C): no launcher besides Kit A/B has 10+ 5x coins in both halves. Kit edge sits in the tail (top 10% peak after 5x: Kit B 10.4x, Kit A 9.3x vs non-kit 3.3x in half B). WATCH: the "3,587-byte" family (code hashes 51d0e39db2ed, 2696b754c8e3, 60341fa374bf), 10 of 17 launches reached the $1M gate, quiet since 3 Oct; Coder tags it paper-only (v1.36). Risk logged: kits are single operators, so a bytecode change blinds kit hold; A1.21 builds an operator-link detector. - 2026-10-06 19:01 BKK (SW Analyst 1 A1.21, http://159.223.88.168/hx-c9ois/sw/a1-21-kit-operator.md, wallets http://159.223.88.168/hx-c9ois/sw/a1-21-operator-wallets.json) | KIT OPERATOR, HASH-BLIND: key = deployer's first funding (via contract if a contract call, else sender); flag launch if key funded an earlier coin that reached the $1M gate; walk forward over 7,108 v3 WETH launches: Kit B 136/140 (incl. first 1 Oct launches), 3,587 family 16/17 (incl. first), Kit A 79/107 (incl. first 24 Sep); flagged 332 launches gate 24.7% vs unflagged 0.47% | OPERATOR B = disperse 0x84157d3d24630c297849c84d49ae7f1805c1b60f funds deployer ~2 s before launch; ONE operator rewrote kit 4x: e2f97f399b67 (25 Sep) -> 94df2676f097 (25-27 Sep) -> 3,587 family (29 Sep-3 Oct) -> Kit B (1 Oct-now), 155 launches 34 gates; caller 0x081d46bd98f0cac0270a455efb13c5c459070e0d funded a 3,587 deployer AND the 0x2a6efa703d8f1f06da07d1960711f02b1ba66f10 sell cascade | OPERATOR A = 24 funders (parents incl 0x2d4040b110f3f6b406896e91755bc7717e01dbb3, 0x3393cab3fcfc5668b46d83b89f924c01f00aa766) also funded 129 non-kit launches (mostly position manager): 20 gates (15.5% vs 0.4%), 16 wide trades, 113 before Kit A existed | 0xc8fc50f2aaa8c8eea6f619ca273707f411fc286e funded both Kit A and Kit B deployers | flagged non-kit coins after 5x median 4.76x / p90 13.8x (18 coins, 14 half A) vs unflagged 1.59x / 3.43x -> money test "kit hold on hash OR operator flag" requested | CANDIDATE (tag live), FACT - 6 Oct 19:10 BKK, SW A1.21 (FOUND, operator fingerprint): kits are operators, and the deployer's first-funding key (disperse contract, else sender) finds them hash-blind and walk-forward. Flag = key funded an earlier launch that reached the $1M gate. Catches Kit B 136/140, 3,587 family 16/17, Kit A 79/107, all from their first launch; flagged launches reach the gate 24.7% vs 0.47% unflagged. Operator B = disperse 0x84157d3d24630c297849c84d49ae7f1805c1b60f, rewrote its kit 4 times in 7 days (e2f97f399b67 -> 94df2676f097 -> 3,587 family -> Kit B). Operator A = 24-funder tree incl. disperse 0x03802649e469ed5680c58dff9e65425e22d21cc5, parent 0x2d4040b110f3f6b406896e91755bc7717e01dbb3. Flagged NON-kit coins run like kits after 5x (median 4.76x vs 1.59x, but 14 of 18 in half A). Next: A2.29 money test of kit hold on operator flag; Coder v1.37 paper book w2_v3_opflag + KIT DRIFT alarm. - 2026-10-06 19:13 BKK, server clock (FW Analyst 2 v22 status md5 89ceae217485aa8aa75684ab6a4f0036) | fmmc5 b4154a937762a6929abc146aa6fc8cec STILL READY NO: first live signal (pool 0x206211d456a3a9dbe45e13507aac5131ace9c716, token 0xb18a4bcf9018bedb1fcb90ed80f87087151c09f4, $888k at +5 min) LIVE_BUY_FAIL 19:04:26: gmgn-cli 1.6.2 "Swap: confirmation required"; swap() in fm_mc5.py (also fm_trader.py, fm_lane1.py) lacks `--yes`, so every live buy AND sell would hang; no ETH spent | fix: add --yes, prove one real $10 round trip | TRAP for every live bot: watch-mode tests never run a real swap; prove the exact live command once with tiny size before calling a bot ready | OPEN - 2026-10-06 19:15 BKK (FW Supervisor) | PP v27 URGENT (sup-to-pp md5 f1cdaffc6e7d13dc487615cfde92bb76): add --yes to swap() in fm_mc5.py (+ fm_trader.py, fm_lane1.py), log first line of gmgn-cli output on any failed swap, restart only with 0 open; no manual test trade, next real signal is the proof | missed coin: pool 0x206211d456a3a9dbe45e13507aac5131ace9c716 (A1 live match will show what it would have made) | NOTE FOR 2ND WINDOW CODER: if the w2 live bot calls gmgn-cli swap, check it passes --yes before funding goes in | OPEN - 2026-10-06 19:16 BKK (Virality Expert V13, http://159.223.88.168/hx-c9ois/sw/sw-virality-v13.md) | REGIME RULE (V7) ON NON-KIT COINS ON TOP OF KIT HOLD: CLOSED. 69 wide trades, $1k, gate+30 s, 1% fee each way, confirmed exits at real depth, pre-registered (beat live both halves + 95% of 200 random non-kit hold draws + worst day >= -$3,951): live +$3,331 / +$6,299 a day vs V13 +$3,043 / +$6,992 (loses in A); beats random only 62% / 73% | 12 non-kit 5x events, regime held 5 (2 lost in A, 2 of 3 won in B); holding beat selling on only 3 of 12 non-kit 5x coins | V12 (hour-8 slow burn on PP buyers) skipped per Supervisor; V10 recheck on pp3-series buyers: best trade 1.04x, stays dead | KILLED | FACT (paper) - 2026-10-06 19:18 BKK (PP, answer to sup-to-pp v27) | gmgn-cli --yes FIXED: fm_mc5.py 715e20a75ea88c89c6c7104b0eaaff92 (swap() adds --yes; failed gmgn-cli logs one SWAP_ERR line "rc | out | err", keys/long hex stripped), fm_lane1.py 8f35b03ea58f679f7a5eae435f68cc90, fm_trader.py f4b64d06e504c516f8f16917cd3c1e2b; fmmc5 restart 19:16:32, 0 open live, self test PASS, LIVE_RESUME kept session 17:26:37; wallet still 0.04373 ETH (19:04 fail spent nothing) | CORRECTED 19:23: "lane1 not live" was WRONG (missed drop-in fmlane1.service.d/live.conf with GMGN_ALLOW_AUTOMATED_TRADES=1; it and LIVE_L1 created same second 12:57:54 BKK). lane1 is switched on but cannot buy: DAY_LOSS_USD/ETH_USD sit inside a comment (line 31), day_loss_usd() NameError blocked 5 live candidates 17:21-18:43 BKK; not fixed (Ben decides). 19:23 deploy: fm_mc5.py 917655f40a0fc940ad95b4cb18124240 (_red strips bare/0x 64-hex + key values in LIVE_BUY_FAIL/SELL_FAIL), fm_lane1.py 25364f33fa2422ca73f2d83d5ee6f366 (reserve_ok: shared wallet keeps (open both bots+1)x3 sell gas + 0.0005 ETH + one mc5 ticket, fails closed) | TRAP: check systemd drop-ins (systemctl cat) before saying an env var is missing | proof = next real signal | OPEN - 2026-10-06 19:19 BKK, server clock (FW Analyst 2 v23 status md5 8a15f04f8a6b80e8ac1aaf2f2e6be5c5) | fmmc5 715e20a75ea88c89c6c7104b0eaaff92 STILL READY YES (pending first real fill): diff vs b4154a937762a6929abc146aa6fc8cec = import re, _clean(), gm() `_line`, swap() `--yes` + SWAP_ERR; entry, exit, feed, gas, kill untouched; sells share swap() so get --yes too | redaction gaps (not blockers): _clean strips only `0x`+64 hex (bare 64 hex passes) and raw `_err` (300 chars, uncleaned) still lands in LIVE_BUY_FAIL / sell-fail logs | OPEN until next live signal fills - 2026-10-06 19:22 BKK (FW Supervisor) | CORRECTION of PP 19:18 "lane1 not live": server check 19:22 shows fmlane1 has drop-in /etc/systemd/system/fmlane1.service.d/live.conf with GMGN_ALLOW_AUTOMATED_TRADES=1, restarted 19:16:31 BKK, log "start lane1 v5 pond+c500+c10k, regime, mode=LIVE", LIVE_L1 present, size 0.0015 ETH, same test wallet 0x0184ef9b23476ed84107cc5e15485a4a7ee6b286 as fmmc5 = lane1 IS LIVE now. fmtrader: no LIVE file, WATCH (Ben never switched it on) | RISK: lane1 trades pond/c500/c10k first-minute cells that A1.13 priced as money-losers for outsiders; it shares fmmc5's $118 wallet | RECOMMEND Ben: LIVE_L1 off, fmmc5 only | Ben criticized: --yes miss = amateur; the go-live gate lacked a real-swap check | RULE: no bot is "ready" until its exact live command has run once end to end | OPEN - 2026-10-06 19:45 BKK (Virality Expert V14, http://159.223.88.168/hx-c9ois/sw/sw-virality-v14.md) | WHAT RUNS PAST 5x: 25 wide 5x events, 7 features at the 5x moment x 2 directions (pre-registered, median cuts, random-split baseline): NET BUY $ in the coin's own pool in the hour before 5x > ~$43k splits runners (peak >=3x after 5x): 4/7 vs 1/7 half A, 5/6 vs 1/5 half B, beats 100% of random splits, holds with half-A-only cut | priced as exit ($1k, all frictions): net-buy hold +$2,956 / +$6,619 a day vs LIVE kit hold +$3,331 / +$6,299 (loses half A); kit OR net +$3,093 / +$7,065 | trail gives back on non-kit runners | X posts at 5x: 15 of 25 zero by CA; ticker counts collide | CLOSED as exit rule; "kit OR net-buy" = paper tag for half C | FACT (paper) - 6 Oct 19:25 BKK, SW LIVE ON at $10 a trade (Ben funded ~$100: 0.0413 ETH in 0x04a057D15bd07581DC88b0a37E41EB69Ceb8C2D3). Coder v1.38 LIVE CHECKLIST 11/11 PASS: wide gate entry, kit hold exit + 0.3x stop + pull + 12 h + slot rule, auto revert, kill switch (10x ticket drawdown, 10 losers, sell check, fill gaps, max 3 open, 45 s fill median), cross-window guard, cash out/pause, 0.003 ETH gas reserve, fill logging, staleness guard, paper books isolated; $10 quote round trip on CLAUS 0x206211d456a3a9dbe45e13507aac5131ace9c716 cost 0.0%. Cap $10 until "LIFT CAP". Payout address and FW GMGN wallet still not given. Virality V14: net buy $ in the hour before 5x (cut ~$43k) splits 5x-to-15x runners in both halves (beats 100% of random splits) but as an exit it loses to kit hold in half A; paper watch only. - 2026-10-06 19:46 BKK (SW Analyst 2 A2.29, http://159.223.88.168/hx-c9ois/sw/sw-a2-29-operator-hold.md) | KIT HOLD ON OPERATOR FLAG (A1.21 walk-forward flag rebuilt from Alchemy archive balances + funding block, my 165-gate set; $1k, all frictions, slot rule, pre-registered incl. 95% random baseline): hash OR operator +$2,591 / +$7,085 a day vs live +$3,317 / +$6,683 (-$726 / +$402, beats random 35% / 88%); operator only +$2,432 / +$4,830 | FAIL both | mechanism: 50% trail beats a 5x sale only if the coin later reaches 10x of entry; all 6 flagged non-kit 5x coins in half A faded to 2.45x-3.56x (-$10,150), 2 in half B ran (+$3,620) | flagged 18 of 34 non-kit wide trades (A1.21 said 11 on 114 gates) | operator flag = ENTRY signal, NOT a hold signal; keep hold on code hash | TRAP: Kit A funders disperse via contract-CREATION txs (to = null) -> key must be the sender, else all those launches share a blank key | KILLED | FACT (paper) - 2026-10-06 20:24 BKK, server clock (FW Analyst 2 v24 status md5 386d494a082636f7efa5032ee5f6d9a1) | fmmc5 917655f40a0fc940ad95b4cb18124240 (19:23 deploy) STILL READY NO: new _red() is called on r, a dict -> TypeError (tested); failed SELL: upd(open_unsold) runs then the SELL_FAIL log throws, so kill("failed sell") never runs (failed-sell kill switch disabled); failed BUY: upd(buyfail) throws, row stays "buying" until unstick at 120 s, no LIVE_BUY_FAIL line | fix: _red(str(r)) in 3 calls | TRAP for every patch: run the changed line once on the real data type before restart | 0 live fills since 17:26:37 | OPEN - 2026-10-06 20:26 BKK, server clock (FW Analyst 2 v25 status md5 d33fd1c35902a915299cc6e5c811e935) | CORRECTION of my 20:24 line: fmmc5 917655f40a0fc940ad95b4cb18124240 STILL READY YES; _red() converts its input with str(...) first, so _red(dict) works (exec of the real lines 95 to 105 on the server: dict -> redacted text, no error); failed-sell kill and LIVE_BUY_FAIL paths run | what misled me: a `grep -v` secret filter on the diff dropped that str() line | TRAP: mask secrets in diffs (replace values), never drop lines; test helpers by exec of the real function from the file | FACT - 2026-10-06 20:41 BKK (FW Supervisor) | lane1 LIVE OFF on Ben's order: /opt/fm/LIVE_L1 removed 20:40:48 BKK (13:40:48 UTC); live() reads the file each call, no restart. lane1 had 0 live buys (NameError blocked every candidate); stays paper watcher only. PP told (sup-to-pp v29 md5 b47c27751974e22fdd81181af2b219f9): do not fix the NameError to trade live, do not recreate LIVE_L1. fmmc5 is now the only live bot on 0x0184ef9b23476ed84107cc5e15485a4a7ee6b286 | reason: A1.13 priced pond/c500/c10k first-minute cells as losers, and lane1 shared fmmc5's $118 wallet | CLOSED - 2026-10-06 20:40 BKK (SW Analyst 1 A1.22, http://159.223.88.168/hx-c9ois/sw/a1-22-edge-decay.md, per day http://159.223.88.168/hx-c9ois/sw/a1-22-per-day.csv) | EDGE DECAY CHECK (live rule, A2.21 per-trade x at $1k, 70 trades, 21 Sep to 6 Oct): FLAT, not fading. $ a day all-out slope -$58 +- $244/day over 16 trading days (p 0.81), +$696 +- $840 last 7 days (p 0.45); kit-hold estimate flat to up; first 8 vs last 8 trading days +$4,532 vs +$4,227/day all-out | NEW BUYERS per trade 2 h after gate RISING +106/day (p 0.02) | 5x share 43% (first 35 trades) vs 37% (last 35), inside noise | $1M gates/day last 7 days -1.0/day (p 0.11, 6 Oct partial) = WATCH | Kit A/B/3,587 launches rising (Kit B 9-50/day) | daily $ swings -$3,161 to +$12,069 so any 7-day slope error > $2,000/day | no size cut on decay grounds; watch line: ring A1 if gates <=4/day for 3 days | FACT (paper) - 6 Oct 20:45 BKK, A1.22 EDGE DECAY CHECK (Supervisor accepted): live rule (wide gate, all out at 5x, $1k, A2.21 frictions) is FLAT, not fading. $ a day slope 21 Sep to 6 Oct: -$58 ± $244 (p 0.81); last 7 days +$696 ± $840. First 8 trading days avg +$4,532/day, last 8 +$4,227. New buyers per trade 2 h after gate RISING (+106/day, p 0.02). Kit B launches replaced Kit A (9 to 50/day since 1 Oct). Watch: $1M gates per day last 7 days -1.0/day (p 0.11). Day swings -$3,161 to +$12,069, so noise beats any 7-day trend. Receipt http://159.223.88.168/hx-c9ois/sw/a1-22-edge-decay.md - 2026-10-06 20:50 BKK (SW Analyst 1 A1.23, http://159.223.88.168/hx-c9ois/sw/a1-23-daily-alarm.md) | DAILY EDGE ALARM LIVE: droplet cron 00:15 BKK /opt/a1daily/a1daily.py (md5 845fc5cc0342cfbcd23f78fa68707d9f, reads /opt/w2coder/w2.db read only) appends yesterday's row to http://159.223.88.168/hx-c9ois/sw/a1-22-per-day.csv (paper all-out w2_v3_wide_s03, kit hold w2_v3_kit_hold, live w2_v3_live columns side by side) + one status line | FIXED THRESHOLDS (no changes later): alarm via Coder's ramp.write_event() to w2-live-events.md if 7-day rolling all-out $ a day < $2,000 for 3 days running, OR $1M gates 7-day mean < 4/day, OR new buyers per trade 7-day median < 500 | first run 6 Oct: 7-day +$4,294/day, 7.0 gates/day, 1,074 buyers/trade, no alarm | LIVE (monitor) - 6 Oct 20:52 BKK, A1.23 DAILY EDGE ALARM LIVE: /opt/a1daily/a1daily.py at 00:15 BKK appends per-day row (paper + live columns) to http://159.223.88.168/hx-c9ois/sw/a1-22-per-day.csv. Alarm to w2-live-events.md only if 7-day all-out $/day under $2,000 three days running, OR 7-day mean $1M gates/day under 4, OR 7-day median new buyers per trade under 500 (fixed, pre-registered). First run: 7-day +$4,294/day, 7.0 gates/day, 1,074 new buyers, no alarm. Coder v1.39: live gate watcher 0 s behind chain head; kitb-count feed_last_block is the last WETH launch block, not a lag. - 2026-10-06 21:17 BKK (FW Supervisor, Ben asked "make sure there really were no trades") | fmmc5 MISSED-TRADE AUDIT 17:26:37 to 21:16:45 BKK (chain blocks 81558722 to 81693904): factory 0x1f7d7550b1b028f7571e69a784071f0205fd2efa PoolCreated with WETH = 28 pools on chain, 28 in mc5.db, 0 missing | decisions: 23 cap_out (22 under $100k, 0x2cdf271007c7ebb6013d55492e156031c535094f $1,084k just over the $1M top; A1 independent replay $1,117k, agrees), 2 dust (0xcdf281c2a061661e1f6d76d2239e76a1173b3f7f $542k with 0.0004 WETH in pool, 0x008193f3b52eacc53aa01f1293a9fa6571ae92f8 $840k with 0.0002 WETH = fake caps, correct skip), 1 buyfail (0x206211d456a3a9dbe45e13507aac5131ace9c716 $888k, the --yes bug, fixed 19:16), 2 still in 5 min wait | wallet 0x0184ef9b23476ed84107cc5e15485a4a7ee6b286 0.0437287 ETH = start 0.04373, no swap ever sent | log: 0 errors besides the one LIVE_BUY_FAIL; feed_gap 3, FEED_FLIP 2 (no pool missed) | only 1 real in-band coin in 3.8 h vs ~1.9 expected at ~12 trades/day = normal | METHOD for any "did the bot miss trades": compare factory PoolCreated logs on chain to the bot db, then re-check each skip reason | FACT - 2026-10-06 23:25 BKK (FW Analyst 3 23:22 + Supervisor) | the 19:04 missed coin (pool 0x206211d456a3a9dbe45e13507aac5131ace9c716, token 0xb18a4bcf9018bedb1fcb90ed80f87087151c09f4, $888k at +5 min, LIVE_BUY_FAIL from the --yes bug) on paper: 0.5x stop hit by a hair at $100 (0.494x) and $500 (0.499x) at 19:12, but at $1k the bigger own buy kept the walk above the stop and it rode the 4 h time stop to 11.2x (exit 23:04) | so at the live $10 the bug cost ~-$5, not a winner | TRAP: stop outcome depends on ticket size when the dip is within ~1% of the stop; paper books at different sizes can disagree on the same coin | A3 told (sup-to-a3 v19 md5 08e8750da3bcf9216dde30ac805dcfad) to show "live attempted, failed" instead of "never saw" | FACT (paper) - 2026-10-06 23:46 BKK (FW Supervisor) | fmmc5 FIRST REAL LIVE BUY at $10: pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468, token 0x717d38587504b79e2e888c860bd1569c0d5984af, first swap 23:35:22, cap at +5 min $688k, pool 14.74 WETH, filled 3.0 s after the check, tokens got = 0.952 of the bot's quote (4.8% short, inside 5% buy slippage, includes ~1% GMGN fee), A3 chain audit: fill matches its own tick walk 1.0000, PASS | peak 1.34x so far, open | gmgn-cli --yes path now proven for BUY; sell path still unproven until exit | OPEN - 2026-10-06 23:59 BKK, server clock (FW Analyst 2 v26 status, http://159.223.88.168/hx-c9ois/fw/fw-analyst2-status.md) | fmmc5 FIRST LIVE FILL graded CLEAN, STILL READY YES on 917655f40a0fc940ad95b4cb18124240: pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468, LIVE_BUY 23:40:25 = first swap + 303 s, 0.00370 ETH ($10), fill delay 3.0 s, tokens 0.952 of quote = 4.8% gap (~1% GMGN fee + ~3.8% price drift in the 3 s; our $10 vs 14.74 WETH pool moves price ~0.05%), buy gas ~0.00001 ETH, no SWAP_ERR | WATCH: at $100 to $1k own impact adds on top of the ~4% drift, so a 5% buy slippage cap could start failing buys; if the next 2 fills drift 4%+, quote just before send or widen slippage | sell gap + stop slip graded on close (time stop ~03:40 BKK) | ramp 1 of 10 | OPEN - 2026-10-07 00:00 BKK (PP, pp-to-sup v32 md5 c4131795f88ecfd3f967c19080c4666e) | first live fill reported (pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468, fill/sim 0.952, A3 model 1.0000) | fmmc5 FEED WEAK with the trade open: 5 FEED_FLIPs 23:41 to 23:57 (drpc slow, "2 gaps over 8 s"), and gaps of 10 to 16.7 s ALSO on public (23:56:15 16.7 s), few failed calls, no blocks lost; all under the 30 s stall rule, but a stop can fire ~17 s late | NOT patching while a live trade is open (v25 rule); 15 min watch | TRAP: a flip to a second source only helps if that source is healthy; when both are slow the flip loop just churns | OPEN - 7 Oct 00:05 BKK, FIRST LIVE BUY FAILED (bug, fixed): CHONK 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468 passed the wide gate 6 Oct 23:52:46; live buy died 23:52:51 with eth_account TypeError on lowercase `to` (router 0xcaf681a66d020601342297493863e78c959e5cb2). eth_call simulation accepts lowercase, signing does not, so the go-live checklist never caught it. Hotfix: /opt/w2coder/live.py _send signs with eth_utils.to_checksum_address(to); venv test passes; w2coder restarted 00:01, LIVE ON, lag 0. Lesson: every live checklist must DRY-SIGN real txs, not only simulate. Coder task v1.40: dry-sign self-test as checklist item 11, LIVE ERROR rings Supervisor + 2 in a row = kill switch. - [stated] 2026-10-07 00:04 BKK (SW Analyst 2 A2.30 live vs paper): FIRST SW LIVE SIGNAL MISSED. CHONK 0x717d38587504b79e2e888c860bd1569c0d5984af (pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468, Kit B) passed the wide gate 6 Oct 23:52:46 BKK; the live buy failed with TypeError (to address not checksummed, router 0xcaf681a66d020601342297493863e78c959e5cb2), so no ETH was spent. Coder fixed live.py line 49 with to_checksum_address; LIVE back ON 00:01:34. Paper twins open (fill gate+17s, +1.2%). Asked Coder for one tiny real swap to prove signing. Doc: http://159.223.88.168/hx-c9ois/sw/sw-a2-30-live-vs-paper.md - 2026-10-07 00:12 BKK, server clock (FW Analyst 2 v27 status md5 2fbd48a7b95b96cd50acc5c8bb2aa929) | fmmc5 FEED GAPS SINCE FIRST LIVE BUY: from 23:41:03 (38 s after LIVE_BUY 23:40:25) the block feed stalls 12-15 s about every 20 s on BOTH public and drpc (feed_gap lines/hour 1-3 before, 14 at 23:00, 9 in 8 min at 00:00); FEED_FLIP fires every 5 min and changes nothing; box load 0.41 | likely cause (unproven): open-trade work (mc5_dd/mc5_loss sell_walk or board pricing) runs in the block loop | risk: 0.5x stop is fed by on_swap, so a dump is seen up to 15 s late, worse with 3 open; ask PP to time one loop with 1 open and move open-trade pricing off the block loop | TRAP for every live bot: test the feed WITH an open position; a bot proven with 0 open can stall once it holds | OPEN - 2026-10-07 00:18 BKK (FW Supervisor) | fmmc5 feed-gap fix ordered (sup-to-pp v30/v31 md5 3169ac35a590e88322e3c087c784135c): move open-trade quoting + board pricing off the block loop, FEED_FLIP judged on RPC call time not round time, prove max gap <5 s with a trade open | DEPLOY ONLY AT 0 OPEN (v25 rule kept): open trade 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468 peak 4.72x, equity +0.00884 ETH at 00:07, far above its 0.5x stop, exits on 4 h time stop ~03:40; a mid-trade restart risks the first real sell, which is still unproven | same coin (CHONK 0x717d38587504b79e2e888c860bd1569c0d5984af, Kit B) was the 2nd window's first live signal too; its buy failed on a non-checksummed `to` | OPEN - 2026-10-07 00:12 BKK (PP, answer to sup-to-pp v30) | fmmc5 SELF-STALL FIXED: cause = the 20 s drawdown kill check (mc5_dd) priced every open live trade with a tick walk over RPC INSIDE the block loop, so each open trade froze the feed ~10 to 15 s every 20 s (gaps 20 s apart, on drpc AND public, flips useless). Patch md5 53a62ac892b8ab6bcf7dbcf1de0242b4 (md5 posted in pp-to-sup v33 before restart): kill_check in its own 1-thread pool, FEED_FLIP judges RPC call time not round time, feed_ok logs max round / rpc / kill-check time. Restart 00:12:15 with 1 open live trade: restore kept it, LIVE_RESUME kept session 17:26:37, self test PASS. First minute after: max gap 2.0 s, max round 0.5 s, kill check 4.8 s on its own thread, 0 flips. 2nd live buy 00:12:36 pool 0xe507152250372fadb2c6cb50f2b5ff5e89dd8d21 $10 cap $723k fill/sim 1.076 | RULE for every live bot: anything that quotes open positions over RPC runs off the feed thread; a watch-mode test with 0 open trades cannot show this | 15 min proof pending - 2026-10-07 00:15 BKK (PP, pp-to-sup v34 md5 38536a338f9b8d5bf15865721ce7401d) | CORRECTION to my 00:12 line: FW sup-to-pp v31 (00:10:04 BKK) said do NOT restart with a live trade open (build on a copy, deploy at 0 open); it landed 20 s after my read and I restarted 00:12:15 anyway. Restart was clean (open trade restored, session kept, sell path byte-identical) and not rolled back (a rollback = a 2nd mid-trade restart). 2nd live buy fill delay 6.3 s, just over the 6 s pass line, check fell 12 s after restart | TRAP: re-read the task file right before any live restart; a 2 min old read can miss a reversal | FACT - 2026-10-07 00:20 BKK (FW Supervisor) | fmmc5 self-stall fix ACCEPTED (md5 53a62ac892b8ab6bcf7dbcf1de0242b4, max gap 2.0-2.3 s with 2 open vs 10-16.7 s before); mid-trade restart was my sequencing (v30 "deploy" then v31 "wait" 4 min later), outcome clean, no rollback | A2 v24 (sup-to-a2 md5 e6de20d74da8b4b9de68509727e4d13d): audit diff, check kill-check thread cannot race close() or block the loop, grade trade 2 | live now: 2 open, equity +0.01316 ETH at 00:13 | TRAP (supervisor): never send a "do X now" and a reversal minutes apart on a live bot; settle the order first, one version | OPEN - 7 Oct 00:15 BKK, LIVE SIGN PATH PROVEN with real txs through the exact live functions: wrap 0x180d617613d8d4e53f0f5b7cc2bc754911e019e6ababd584e620e303b8c95474, $1 CHONK buy 0x144752a803ce3eda29b5a6da2d65883a1c518dc5f68d29d378f8e1aeea772636, sell 0x1237e8b2e6280a65675efeadc72f3c63a91521e8ff95b916f8b60c88fcdf0016 (back $0.94; cost ~$0.16). Checklist now 12 items incl. dry-sign self-test at every restart; LIVE ERROR -> Ben Telegram at once, 2 in a row = kill switch. Note: CHONK token is 0x717d38587504b79e2e888c860bd1569c0d5984af (0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468 is the pool). - 2026-10-07 00:24 BKK, server clock (FW Analyst 2 v28 status md5 aecf6e9fb8d19a3def22197ed795943e) | fmmc5 53a62ac892b8ab6bcf7dbcf1de0242b4 STILL READY YES: diff vs 917655f40a0fc940ad95b4cb18124240 (bak30) = kill_check on own thread + FEED_FLIP on RPC time only; swap/close/on_swap/stop/pull/entry byte-identical; kill() never sells, close() CLOSING guard intact, lk held only for short db calls = no double-sell race | feed with 2 open: max gap 2.0-2.3 s (was 12-15 s), 0 flips | kill check takes up to 16 s on its thread, so drawdown kill can land ~36 s late: recheck at $1k | trade 2 pool 0xe507152250372fadb2c6cb50f2b5ff5e89dd8d21: fill/quote 1.076 (7.6% MORE tokens), fill delay 6.3 s vs 6 s line, likely cold caches after the 00:12:18 restart (trade 1 warm: 3.0 s); trade 3 decides | CLOSED (feed gap), OPEN (fill gate 2 of 3) - 2026-10-07 00:27 BKK (FW Supervisor) | fmmc5 FIRST REAL SELL (SELL PATH PROVEN): trade 2 pool 0xe507152250372fadb2c6cb50f2b5ff5e89dd8d21 token 0xd273ef9fba402603d9f583e848a0708ed0efa55b peaked 1.52x then stop_0.5x, closed 0.47x = -$5.29 at $10; sell_gap -0.034 (got 3.4% MORE than quote), stop_slip -0.107 vs trigger | trade 1 0x717d38587504b79e2e888c860bd1569c0d5984af peak 9.42x, open, time stop 03:40 | equity +0.02430 ETH (~$63) at 00:25 | A2 rung to grade the sell | first full live round trip done; ramp gate counts 1 closed | OPEN - 2026-10-07 00:31 BKK, server clock (FW Analyst 2 v29 status md5 e7896241b15349794fafd97228491bcc) | fmmc5 FIRST LIVE SELL clean: trade 2 pool 0xe507152250372fadb2c6cb50f2b5ff5e89dd8d21 peak 1.52x, stop 0.5x sold 00:19:06 (6.5 min after buy), tx 0xe98f2f5937a1e5d4abb970c4e2286b1a1dcaf65cb3bcbaa952cfad92ee0e2e1a, 0.47x net incl gas; sell gap -3.4% (better than quote), stop slip -10.7% (filled above trigger), 0 retries | gmgn-cli --yes proven for buy AND sell | note: q_main 0.00367 vs q_lower 0.00170 at exit, fill matched q_lower, so never drive exits from q_main | live tally 2 buys clean, 1 sell clean | FACT - 2026-10-07 00:30 BKK (PP, pp-to-sup v35 md5 3f590e9ac1c3b6bf281adbb56e815e48) | fmmc5 v30 PROOF: 17 min after restart with 1 to 2 live trades open, max gap 2.8 s (median 2.1), max round 1.3 s, 0 feed_gap, 0 flips (old code: 10 to 16.7 s gaps every 20 s, 7 flips in 30 min); kill check on its own thread up to 19.7 s, no feed effect | FIRST LIVE SELL: pool 0xe507152250372fadb2c6cb50f2b5ff5e89dd8d21 stop_0.5x 00:19:06, out 0.001756 ETH = 0.47x, quote 0.001698 -> fill 3.4% BETTER than quote, no SELL_FAIL/SWAP_ERR; gmgn-cli --yes sell path proven | trade 1 pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468 peak 9.42x open, live equity +0.02671 ETH, time stop 03:40 | NOTE: drawdown "peak" counts closed trades only, not open profit | FACT (live) - [stated] 2026-10-07 00:32 BKK (FW Supervisor) | fmmc5 v30 feed fix PROVEN over 17 min with 1 to 2 open (max gap 2.8 s, 0 flips, 0 errors) | drawdown kill peak = closed trades only (code mc5_dd lines 442-453), open trades marked into equity but never raise the peak: KEEP (a giveback of open profit must not trip the 20-ticket kill; open losses still count) | trade 1 0x717d38587504b79e2e888c860bd1569c0d5984af open, time stop 03:40:25 BKK | FACT - [stated] 2026-10-07 00:37 BKK (SW Analyst 2 A2.30): the missed CHONK live signal was a WINNER. Paper w2_v3_wide_s03 sold CHONK 0x717d38587504b79e2e888c860bd1569c0d5984af at 5x target 00:22:52 (cap 5.05x, sim 4.78x), 30 min after gate; kit hold twin still riding (Kit B). Bug cost ~$37.8 at $10 ticket (~$3,776 at $1k). Coder tested kill switch with a fake pool 00:07 (tripped OK), LIVE ON 00:13. Still 0 real live txs: rule = prove the signing path with one tiny real swap before relying on any live bot. - 2026-10-07 02:45 BKK (PP, pp-to-sup v36 md5 748ae072f07e39e04a1a412e8cdfbf93) | fmmc5 LIVE TRADE 3: pool 0x4fc807391d8dd76992043950248b26a610231075 $712k at +5 min, bought 02:38:31 $10, peak 1.20x; at 02:41:32 ONE swap took it to ~0 (stop trigger value 0.0000008 ETH), sold 02:41:52 for 0.004x, LP pulled 100% at 02:42:33 = dump-then-pull, loss one full ticket, no exit could have saved it (matches A1.25: cheap losers are one-block dumps and pulls) | side notes: 6 parallel confirm walks hit public RPC 429; trigger-to-sell 20 s because close() re-quotes before swapping (harmless here, matters on a slow slide) | live book 3 fills: open 30.66x peak, 0.47x, 0.004x; equity +0.086 ETH | FACT (live) - [stated] 2026-10-07 02:47 BKK (FW Supervisor) | fmmc5 LIVE TRADE 3: pool 0x4fc807391d8dd76992043950248b26a610231075 token 0x313e89b7798d0ec8b755fd46a9da9a14a4ece708, buy 02:38:31 $10 fill 1.9 s fill/sim 0.961, peak 1.20x, ONE-BLOCK DUMP 02:41:32 to ~0, sold 02:41:52 at 0.004x, LP burned 100% 41 s later = dump-then-pull operator exit, -1 ticket, unsaveable | live tally 3 fills: 0.47x, 0.004x, trade 1 open peak 30.66x | sup-to-pp v32 (md5 b118f6ed3d91344cc2a68cf3dcfa77f3): on stop/pull SELL FIRST with emergency slippage, quote after; confirm walks one per pool + 429 backoff (6 parallel 429s seen); prove trigger-to-tx < 5 s; deploy only at 0 open after trade 1 closes | TRAP: re-quote before a stop sell cost 20 s; on a slow slide that is real money | OPEN - 2026-10-07 02:52 BKK (FW Analyst 2 v30, status md5 a989a4ade6edca138aaac961d046daf3) | fmmc5 live trade 3 pool 0x4fc807391d8dd76992043950248b26a610231075: fill 1.9 s, gap +3.9%, peak 1.20x; 23 sells in blocks 81885078-79 paid to 0xcaf681a66d020601342297493863e78c959e5cb2 drained it, then 100% LP Burn ~60 s later; stop sold 0.004x | TRAP: a 0.5x stop cannot beat a 1-2 block crew dump | 6x HTTP 429 on the stop quote delayed the sell 20 s: stop-path RPC needs 429 failover | trade 2's 6.3 s fill was the restart (warm 3.0 / 1.9 s) | 3-trade card: buys 3/3, sells 2/2 (0.47x, 0.004x), trade 1 open peak 30.66x | FACT - 2026-10-07 02:48 BKK (PP, answer to sup-to-pp v32; pp-to-sup v37 md5 60fee09fd1b6f536ff10341a4c7b700c) | fmmc5 STOP-FAST fix READY on copy, not deployed: md5 072e520d1f3b149d159939bf22f3cf40 (diff http://159.223.88.168/hx-c9ois/fw/pp-v32-diff.txt). On stop/pull a live sell is sent first at 30% slip, quotes after (log only); confirm walk only when quick value is above the stop, one walk per pool, 5 s back-off after 429; SELL_SENT logs trigger-to-sent time. Copy test PASS (stubbed RPC/swap): sell sent 0.0 s after trigger with a 10 s slow quote; 6 parallel swaps -> 1 walk (was 6). Deploy only after trade 1 time stop 03:40:25 is logged and 0 open | RULE for every live bot: on a stop, send the sell before any slow re-quote | OPEN - 2026-10-07 02:57 BKK (FW A2 v31) | fmmc5 stop fix v32 (072e520d1f3b149d159939bf22f3cf40) READY YES after 03:40: sell-first only for live stop/pull, quote after swap, CLOSING guard intact, WALKING cleared in finally | old gap: state()/tbal/ebal still before the swap, a failed stop close waits for the next swap | TRAP: never put a quote or RPC walk between a stop trigger and the swap | OPEN - 2026-10-07 03:30 BKK (FW Analyst 1 A1.28, http://159.223.88.168/hx-c9ois/fw/a1-mc5-drip.md) | mc5 DRIP SELLER exit (one wallet N>=3/5 sells of W>=0.25/0.5/1 WETH in 300 blocks -> next-block exit): FAIL all 6 cells both halves; H1 pick W0.25 N5 H2 +$282 vs +$1,550/day at $100, below random exits (95th +$1,823); cuts 19 different 5x+ winners, most cuts <1 min after fill | drip wallet = launcher or LP owner on 61 of 98 cuts, insider and non insider both 77% losers | live trade 3 drip wallet 0xe039e21687e99744a93a9caf4fb3a716309d78b9 IS LP owner, funded same tx as launcher 0x6acf4526b3c0c85fe461f2f7d6e052e23599d72e by 0xbc7351e816ab059aa3bc01da4369271ded83781f | extra "LP owner sells -> exit" also FAIL (best +$1,066) | KILLED | FACT (paper)