---
name: trading-edge-hub-7
description: Trading edge hub ledger part 7 (continues trading-edge-hub-6, which is at its cap): from 6 Oct 2026 ~11:30 BKK. Read with [[trading-edge-hub]] before any buy/sell call.
sources: [cowork]
---
# Trading edge hub, part 7 (follows [[trading-edge-hub-6]])

## 6 Oct 2026 ~11:45 BKK: wash fix + creator rule (Price Predict session)
- [stated] Ben: set up servers for full power if needed; do the wash fix rerun, the creator rule backtest and other priorities.
- 2026-10-06 | WASH FIX ACCEPTANCE PASSES (20 Sep..3 Oct, lab swaps with known trader, first-funder army rule): crew A 0x92a0849f9b1e9e02363a4bd8d74a317fa6ddae92 wash 97.7% by count / 100% buy ETH / outside 5.0 ETH / outside 0.1+ ETH buys pulled <60 s 2 of 2 / army 5.9%; crews B/C 0x4757fc41278bd0f1b028c6dae71d381dc4558d98 96.8% / 100% / 5.9 ETH / 5 of 5 / 7.9% (old repeat rule 47.5% / 67.0%). Army 1,370 wallets from 0xbae538011874261e3fb4ca6aa03cacb2766c4d81, 1,078 from 0x79411dbcbcf0e5b27cd222f24e082fc0bbf32f16. Gap: swaps after 30 min have no trader wallet in day files (35-46% of these crews' swaps), left out | doc http://159.223.88.168/hx-c9ois/brx/pp-wash-fix-result.md (md5 0c272122102252cfe1f93858b6090490), project claude/pp-wash-fix-result-2026-10-06.md | FACT. CORRECTS my 09:15 line and the lab 07:00 point (6): the pull trigger after outside 0.1+ ETH buys is real for A and B/C (7 such buys in 14 days, all pulled within 60 s).
- 2026-10-06 | wash.py patched (backup wash.py.bak-ffund-*): army = repeat OR first funded by a distributor (seed + any funder of 20+ wallets trading one crew with 50%+ share; /opt/pricepredict-data/distributors.json), ETH/WETH pools only (fixes 1.9B ETH quote bug), label CHECK if outside ETH > 10,000. First funders: new /opt/pricepredict-data/ffund.py (systemd unit ffund, 48 threads, nice 8, paid dRPC, ffund.db, 181,065 crew-coin wallets, A/BC first then by buy ETH; block trace files sibling wallets free). One-shot rerun washfix-rerun.timer 13:58 BKK + nightly cron | LIVE
- 2026-10-06 | CREATOR RULE BACKTEST (lab honest fills, entry 10 s, half 3x rest 10x, $20): standard launcher token 3854_17ad80 exists only from 30 Sep (0 pass coins 20-29 Sep). 30 Sep..4 Oct pass n389 3x 46% 10x 18% +59%/coin (review +60% n392 matches); fail +4% overall but -85%/-81% on 3-4 Oct and +92%/+145% on 30 Sep/1 Oct (fail crew 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 +440% n32). Pass = 6 crews, fail = 23 crews, zero overlap: fingerprint = crew badge. Pass inside crews A+B/C n234 +99%/coin every day positive (+32..+156%); other 4 pass crews n155 -2% | doc http://159.223.88.168/hx-c9ois/brx/creator-rule-bt.md (md5 68af0b317d12bb1ae0a5259325e2fde4), project claude/creator-rule-bt-2026-10-06.md | FACT (5 days). Best rule: pass AND crew A or B/C.
- 2026-10-06 | Dune key over its credit limit for this billing cycle (upload refused 402).
- 2026-10-06 ~12:10 BKK | [stated] Ben created DigitalOcean droplet "research-box" 152.42.241.219 (private 10.104.0.5; SGP1, 8 vCPU / 16 GB / 320 GB, $96/month, key kookwatch-deploy; main reaches it with /opt/pbsync/pbkey). Purpose ([stated]): carry the processing load the 2 to 4 Cowork sessions already run. Copied: /opt/hist/rh day files, crew.db, wash.db, drpc_paid.json + twitterapi_key (chmod 600). ffund.py moved there (~1,350 wallets/min vs ~300 on main), watcher + snapshot cron there; main pulls ffund_snap.db every 10 min (pull_research.sh) for wash.py; main pushes wash.db nightly 00:40 UTC | LIVE
- 2026-10-06 12:30 BKK | C10K VERIFIED on PP fills (lab real mode, entry 5 s, 2 s lag through depth, $20, sell all 3x, review HOT rule, review c10000_200 pass list 25 Sep..4 Oct): 3 ETH HOT n3,051 3x 64% +82%/coin every day positive (+42..+103%); 5 ETH HOT n1,317 +82% (25 Sep n2 -100%, rest positive); COLD 3 ETH +5%, 5 ETH -19% (review +111/+90, COLD +36/-22). c10k = 710 crew roots (0x040571e5c4d2efdde0901656f1dba0e545cca85e 811 coins), none run pond/c500 | doc http://159.223.88.168/hx-c9ois/brx/pp-answers-6oct.md (md5 fd31d069108e2c1d677a441eaae65d54) | FACT
- 2026-10-06 | BADGE + CREW WALK FORWARD (supervisor push back accepted: A/BC were picked after the fact): crews chosen only from earlier days, test 1..4 Oct: all badge +51% (n351), crew mean>0 +99% (n218), crew 3x>=50% +122% (n158), better on every test day | FACT (4 test days). Launch-time tag: first-4 s buyers are the same ~26 bots in every badge coin (0x2be87ad70cf11ea294d7c42044b5b8277a3e4874 in 386 of 392) = useless; 71% of badge creators were funded via 0xf341c935599822f1bc1ec29e27606724fbbf6595 by an EARLIER creator of the same crew, no funder fed two crews = creator-funder lookup can tag crew at second 0. Files: brx/pond_crew_labels.csv (md5 8ebe0d2bd96f73b2e7231f7c2dc7e4db), brx/crew_wallets.json, brx/crew_lookup.json (19,264 wallets) | LIVE
- [stated] 2026-10-06 Ben forwarded sw-tasks-pp.md (PP1..PP5) to Price Predict. PP1 v3live.py deployed on research-box (systemd v3live, /opt/sw/v3live.db: pools, swaps with signer/usd/mcap/pool ETH, liq with share removed) | IN PROGRESS
- 2026-10-06 11:55 BKK | wash gap fix: day files keep token transfers only 30 min, so later swaps had no wallet; wash.py now fills them with the tx signer from /opt/pricepredict-data/signer.db (signer.py: eth_getBlockByNumber full, 25 blocks per request; wash.py writes signer_need.txt), backup wash.py.bak-presig. washfix-rerun2.timer 14:01 BKK runs wash -> signer -> wash -> crewgrade; nightly cron now wash -> signer -> crewgrade. ffund.py sped up 6x (batched 8-ary balance search; dRPC caps us at ~50 HTTP requests/s, calls are ~30 ms): ~300 wallets/min; watcher cron */10 ffund_watch.sh restarts it; nightly 21:50 UTC rm ffund.done = incremental run for new wallets. pp-wash-fix-result.md md5 now 5d41da37a3598cbee0847c9a06da06cb | LIVE

## 6 Oct 2026 ~12:00 BKK: second window, v3 runner rule (phase 2 session; http://159.223.88.168/hx-c9ois/hx/second-window-v3.md md5 67a89149d91af7cb9a5bc1511a437f37, project claude/second-window-v3-2026-10-06.md)
- [stated] Ben: 10 iterations to find patterns of multi-million runners and the best second-window buy (window 1 = seconds after launch; window 2 = when a time, price range or pattern shows big-runner potential); focus Robinhood Chain, use PricePredict data.
- 2026-10-06 | where sellable $5M+ runners come from 13 Sep..6 Oct (own RPC pull: every Swap/Mint/Burn of 7,718 v3 WETH/USDG pools + every Swap of 2,243 Pons graduated pools; sellable = $20k buys at >= level) | Pons graduations (879 new coins): $1M 57, $5M 5, $20M 1 (AGE); Uniswap v3 (factory 0x1f7d7550b1b028f7571e69a784071f0205fd2efa, mostly WETH 0.01%; 926 new coins): $1M 165, $5M 60, $20M 24, ~2-3 $5M coins/day since 22 Sep, path $1M <1 h, $5M at 2-7 h, LP pulled 5-20 h after start (78% pulled); SC, SI, HI, VRAX in this lane; 878 distinct deployers for 926 coins (fresh wallets) | /tmp/w2/an v3build.py, base.py (container) | FACT
- 2026-10-06 | blind buy at first sellable level, $500, sell 5x / stop -40% / pull exit / 12 h, 1% fee each way, depth impact from pool virtual reserve, 5 s fill, halves to 27 Sep / after | v3 $500k 1.01/0.95x, $1M 1.44/1.30x, $2M 1.14/1.50x; Pons 0.7-1.0x (no edge); v3 $1M robust to delay 5 s..15 min (1.13-1.44x) and size $200..$5k (1.62..1.08x); median 0.57x, ~23% hit 5x | sim2.py | FACT
- 2026-10-06 | V3 RUNNER RULE: first sellable $1M AND age <= 1.5 h from first trade AND top receiver share of buys <= 8% AND army share ~0 (buy volume from wallets that bought early in 3+ earlier coins, walk-forward) | 1.86x (n28, to 27 Sep) / 1.90x (n31, after), win 32-39%, best 3 removed 1.72x; $500 a trade +$26k over 17 days, 13 of 17 days green, worst day -$1.4k; 1-6 Oct n20 2.1x; $2k 1.72-1.77x, $5k 1.50-1.55x; 60 s delay 1.75x; every exit 2x..10x / stop 30-50% > 1.19x, best 5x-8x | feat.py, wal.py (1.7M pre-gate transfers, 300 coins) | LEAD (cuts set on first half; small n; crew-made coins; token sell blocks not checked on chain). Asked PP for paper book w2_v3_1m | PROPOSED
- 2026-10-06 | wallet signs at v3 gate, both halves: army share high = bad (0.93/0.88x at $1M); many brand-new buyers = good; holders/top10 weak | FACT

## 6 Oct 2026: 2nd Window Supervisor session
- [stated] Ben set this session as 2nd Window Supervisor: make the second window buy strategy as profitable as possible; team = 2 analysts + 1 Code Owner (2nd Window Coder); no code yet, discovery phase.
- 2026-10-06 | exit math for "half at 3x, rest 10x" when the crew pulls after 3x: a 3x hit with no 10x returns 1.5x, so breakeven needs 3x hit rate >= 66.7% before fees. Explains crew 0x0ec66d31aab7e387b2f4cda075e39f7a6a0f541b at 8 of 14 hits = 0.86x (analyst live: down 10%) | FACT (arithmetic). Fix = per crew exit ladder from that crew's pull level distribution
- 2026-10-06 | supervisor rulings: "pass AND crew A or B/C" NOT live until (1) walk forward crew pick (crews chosen on earlier days only) is positive on every test day, (2) a launch time crew tag (first funder of deployer and first block buyers vs distributors.json / ffund.db) matches the 33 s label >= 80%, OR an entry at 35 s still pays | OPEN
- 2026-10-06 | supervisor ruling on V3 RUNNER RULE (w2_v3_1m): priority 1 over the crew badge rule (two halves, n59, robust to exits/delay/size). Math check: 1/3 at 5x + 2/3 at 0.57x = 2.05x before fees, matches 1.86-1.90x. Paper sd per trade ~2.1x, so ~45 trades (14 days) puts breakeven ~3 standard errors away: 14 days is enough. Open risks: (1) crews A/B/C pulled within 60 s after every outside buy >= 0.1 ETH (7 of 7) and $500 is ~0.18 ETH, so check overlap of rule coins with crews A and B/C and test split buys under 0.1 ETH; (2) sell block check (simulate the sell before buying) is mandatory in code; (3) count how many losers ended near 0 (one-tx pull) instead of -43% | PROPOSED
- 2026-10-06 | supervisor proposal (not yet approved by Ben): paper book w2_v3_1m to start now; tiny live $100 trades once sell simulation is coded and first 15 paper trades track the backtest; $500 at day 14 only if positive without the best coin
- 2026-10-06 | supervisor read of VIRALITY MAP v1 (http://159.223.88.168/hx-c9ois/brx/virality-map.md): no look ahead at entry (step 3 uses hours 2 to 4, entry at hour 4). Same mechanism as the v3 rule army filter (fresh buyers good, recycled trench wallets bad), found independently on a different sample = cross confirmation. Ruling: ONE wallet score chain wide (repeat trench share) feeds both rules, built once by PP. Use v3 rule as ENTRY (about $1M, under 1.5 h, 1.9x) and the ladder as the HOLD manager (keep a slice past 5x while steps 3 and 4 hold, sell on fail). Conflicts to test: w2_viral "only unpullable pools" would drop most v3 coins (78% pulled 5 to 20 h); do $20M v3 winners also get pulled? Steps 1+2 edge decaying (+13% all, +6% after 20 Sep) | PROPOSED
- [stated] Ben, 6 Oct: supervisor prioritises and writes tasks; data collection goes to Price Predict, second window strategy and execution code goes to the 2nd Window Coder, analysis to the two analysts; deliver as summary plus one link per role, Ben's own tasks at the end.
- 2026-10-06 11:58 BKK | second window task lists v1 published (priority: 1 v3 runner rule, 2 virality ladder as hold manager, 3 crew badge rule): PP http://159.223.88.168/hx-c9ois/sw/sw-tasks-pp.md (md5 fbd622a61933276b65dfe3b5a07de30a), Coder http://159.223.88.168/hx-c9ois/sw/sw-tasks-coder.md (cd4ad845294f12464e1eb280517c3df2), Analyst 1 http://159.223.88.168/hx-c9ois/sw/sw-tasks-analyst1.md (72bc22cad22389f96bf5e2755d16fa69), Analyst 2 http://159.223.88.168/hx-c9ois/sw/sw-tasks-analyst2.md (13e50c638d37da4cd48d4977d863ea17); status files go in the same /sw/ folder | LIVE
- [stated] Ben, 6 Oct: avoid Dune (expensive); a third analyst (Analyst 3) is available for second window work.
- 2026-10-06 12:10 BKK | Analyst 3 tasks v1 (independent rebuild of w2_v3_1m, sell block census, PULL EARLY WARNING from LP owner/deployer moves before the pull, bad day risk): http://159.223.88.168/hx-c9ois/sw/sw-tasks-analyst3.md (md5 4762df3783afc84b48eee85bb37d1818). PP list v1.1 adds no Dune (md5 74fb4900997b8709710be206fde29ab9). Publish route: Mac ssh key ~/kookwatch/id_kookwatch to root@159.223.88.168, files in /opt/launchwatch/pub/hx-c9ois/sw | LIVE
- 2026-10-06 12:15 BKK | ANALYST 2 HAND-BACK (A2.1-A2.5; status http://159.223.88.168/hx-c9ois/sw/sw-analyst2-status.md, spec sw-a2-1-spec.md md5 f02bd835bc7d8edfbe29e58222597782, results sw-a2-results.md md5 a6db178d22e4716d8b536e002f1b3096, 59 trades sw-a2-trades59.csv, hubs sw-a2-hubs.txt 37 addresses) | spec: top receiver share = buy COUNT share of the v3 Swap recipient (no exclusions) <= 0.08; army share = token amount received from pool/hubs by wallets seen in 3+ earlier study-coin windows <= 0.009; age from pool's first Swap <= 1.54 h; entry first pool event >= gate +5 s; pull exit R <= 0.5 x R at entry; live fixes: use the pool where the gate fires (backtest picked max-volume pool = look-ahead), read supply at gate | losers: 38 of 59, 28 at ~0.57x, 3 at 0.1-0.5x, 7 near 0 (4 crashes through stop, 3 pulls); stop fills 30% worse 1.80x; every loser = 0 still 1.59x | nearby cutoffs: 72 of 72 cells (gate $750k/$1M/$1.5M, age 1/1.5/2 h, top 6-12%, army 0.009/0.02) > 1x both halves (1.39x..2.43x) | hold slice past 5x with buyer ladder (floor 3x): 100% 1.85/1.90x, 70% 1.76/2.05x, 50% 1.70/2.15x = halves disagree, keep 100% at 5x, 70/30 paper only; floor at 5x level ends at once | $20M v3 coins: 19 of 24 pulled for good, median 15 h after $20M, price before pull median 1% of top (crash first, pull late); 3 pulled near top (0x2627cbc6880fcc356a49ce70aee59e18baa46224, 0x60042b65a0c0a90ba75d17879850b6f99d9422eb, HI 0xee049979960d5e42b389b941ec45790ca7b6e1f7) | /tmp/w2/an sim3.py, grid.py, ladder2.py | FACT (paper)
- 2026-10-06 ~12:30 BKK | supervisor read of A2 spec: 3 MORE look-ahead leaks beyond the 2 Analyst 2 listed, all in army share (spec section 7): (1) earlier coins' wallet windows run to their own gate +10 min / first swap +6 h, so can include transfers after THIS coin's gate; (2) study set = coins that reached sellable $500k, which can become true after this gate; (3) 37 hubs chosen from all 300 coins incl. later ones. New task A2.8: rerun 59 + grid with all live fixes at once (pool at gate, supply at gate block, memory/hubs only from data before gate); rule stands if mean stays above ~1.5x. Loser split (7 of 38 near zero) means pull warning worth only ~+0.05x; Analyst 3 A3.3 refocused to EARLY FAIL EXIT (first 15-60 min after entry, 19 targets vs 35 stops). Coder: build C1 from spec with live choices, keep 100% at 5x, C5 = 70/30 with 3x floor paper only. Docs v1.1: analyst2 md5 8600d824ebdc8c91e96daf51960811d5, analyst3 998d476dd70cb3f9abc52f85344d8761, coder 4edabf65de0abd25e2c9a56f976cd2ca | OPEN

## 6 Oct 2026 ~14:40 BKK: Analyst 1 hand-back (crew badge / window 1; status http://159.223.88.168/hx-c9ois/sw/sw-analyst1-status.md md5 dfddf0673bcf60ad1f422c7ce0ded53d)
- [stated] Ben forwarded sw-tasks-analyst1.md to this session (Analyst 1).
- 2026-10-06 | A1.1 FAIR CREW PICK (pond 30 Sep..4 Oct, 537 coins, entry 10 s, half 3x rest 10x, 3 s sell delay, fees, $20) | HOT (crew's last 10 coins already ended before launch, >=5 hit 3x) +169%/coin n189 every day +115..+224%; day-ahead crew pick +147%; pass + crew A/BC +105%; all pond +45% | /root/a1 a11.py, a12.py | LEAD (5 days, fills without depth curve run ~15 points above PP's)
- 2026-10-06 | A1.2 entry 35 s still pays every day: pass+A/BC +47% (+32..+102%/day), hot +99% | FACT (5 days)
- 2026-10-06 | A1.3 per-crew pull profile: crew A 0x92a0849f9b1e9e02363a4bd8d74a317fa6ddae92 median pull 284 s, best all at 10x (+217% fit / +189% test); 0x0ec66d31aab7e387b2f4cda075e39f7a6a0f541b = review crew 0xae3d565892c79fa8fc9243b380283dd419e28ec9, pull 57 s, peak 2.6x, sell all 3x or skip; rule: median pull <90 s all at 2x, 90-200 s half 3x/rest 10x, >200 s all at 10x | LEAD
- 2026-10-06 | A1.4 launch tag (deployer's funder within 3 hops = creator of an EARLIER launch -> that crew; crew.db wallet.funder): A/BC recall 181/234 (77%), 0 false tags, tagged correct 100%; trade on tag at 10 s +145%/coin vs 35 s label +47% | FACT (5 days); live risk = funder row latency
- 2026-10-06 | A1.5 v3 lane (4,611 v3 WETH 0.01% pools, 114 sellable $1M within 1.5 h; 57 of Analyst 2's 59 trades inside): no deployer in crew map, A/BC army ~0% of buy ETH; 0.1+ ETH buys followed by pull within 60 s 0.10% (136 of 137,833 on rule coins) vs random 60 s window 0.37% = big buys do not trigger v3 pulls | /root/a1/v3 | FACT
- 2026-10-06 ~14:50 BKK | supervisor ruling on Analyst 1: window 1 (hot crew + launch tag) promoted to run in PARALLEL with the v3 rule, not priority 3. Hot filter accepted (rule fixed before test, uses only ended coins). Launch tag gate passed on precision (100%, a wrong tag costs money, a miss only falls back to 35 s). Per-crew best-fit exits REJECTED (2 test days, 7-50 coins); use the general median-pull-time rule, walk-forward. A2.7 split buys CLOSED by A1.5. New: A1.6 pull rate vs outside buy size (0.005..0.1 ETH) to find max safe ticket (biggest window 1 lever), A1.7 rerun on PP lab real-mode depth fills, A1.8 walk-forward pull-time exit rule; PP6 pre-tag wallets funded by crew creators BEFORE they deploy (funding precedes launch, so tag = table read, no latency); Coder C8 paper book w1_hot ($20, 10 s on tag else 35 s on label, pull-time exits). Docs: analyst1 v1.1 md5 6d1423d785a5e2d1eed02a0ef592e3b6, analyst2 v1.2 0a90f875c50a1fe10e045eea75f22d64, pp v1.2 3eaa5c7ae5a636b8422ed7dbe599e811, coder v1.2 26a0278cb29bd89dc83bf4750dc32f93 | OPEN
- 2026-10-06 ~15:10 BKK | ANALYST 2 A2.8 STRICT RERUN (http://159.223.88.168/hx-c9ois/sw/sw-a2-8-strict.md md5 ee77fec56b869e56f66212b8f50340a7; trades sw-a2-8-trades.csv md5 9d01f8614ef20f6a3714abc44937a6e8 = new C1 replay target) | all live fixes at once: pool = earliest gate across the token's v3 pools (same pool for all trades), supply at time t from every mint/burn Transfer (986 tokens), army memory + study set + hubs only from coins whose sellable $500k came before this gate and only addresses seen before the gate block (hub = in >=20% of earlier windows, min 10 coins; median 5 hubs/coin), ETH-USD = last hourly close | 58 trades, 1.90x (first half 1.91x n27, second 1.90x n31) vs 1.88x n59 before; dropped 0xe60c1f5d9ba7f62a392a78472a3ab83dd62467a3 (strict army 0.024, was a 0.535x stop), 4 trades moved <=0.05x by the ETH price step; reasons target 19 / stop 34 / pull 3 / time 2; grid 72 of 72 cells >1x both halves (1.38x..2.61x) | /tmp/w2/an s8.py, /tmp/w2/s8/mb.py | FACT: RULE STANDS (paper)
- 2026-10-06 ~15:20 BKK | supervisor accepted A2.8 (no look-ahead left; 1.90x both halves). Coder v1.3: C1 replay target = sw-a2-8-trades.csv, spec section 7 replaced by A2.8 points 3-5 (md5 64f77e2d0aa8fc915394541ffa5c2f97). Analyst 2 v1.3 new tasks while A2.6 waits: A2.9 more trades at same edge (age 1.5-3 h, re-entry after stop with fresh buyers, USDG pools), A2.10 dollar profit per trade by size $250..$5k with worst day / worst 3 days (table for Ben's sizing call), A2.11 daily live parity check vs C1 book (md5 8048645c856c0bb194fc874fa43d7013) | OPEN

## 6 Oct 2026 ~12:45 BKK: team restructure (Price Predict session)
- [stated] Ben: Price Predict owns the code and most processes; PricePredict Supervisor reviews, directs, prioritises and tasks PP; three new First Window Analyst sessions (1, 2, 3) focus on buying in the first minute (current focus). Updates to Ben = one link per recipient (Analyst 1/2/3, Supervisor) plus what Ben does now.
- 2026-10-06 | First window briefs published: shared setup http://159.223.88.168/hx-c9ois/fw/fw-common.md (md5 d808dc7943fd5f211d36462835009692), Analyst 1 pond crews fw-analyst1.md (3874c20a7cf66dbc73501867ebba6ad3), Analyst 2 c10k fw-analyst2.md (a78b2dd6fbbaa5a7770c1e7a67d55b95), Analyst 3 c500 + paper vs real + whole book fw-analyst3.md (9e409fd4ff6c7496f07fe650d8a19619), Supervisor fw-supervisor.md (020e4cf037336ae6af826112874e2d84); status files go to /fw/fw-analystN-status.md | LIVE
- 2026-10-06 | PP6 PRE-TAG LIVE on research-box: /opt/sw/pretag.py (systemd pretag; seeds = creators, creator funders, distributors from crew_wallets.json, 21,964 wallets; every tx FROM a tagged wallet that moves ETH is traced, recipients tagged same crew hop+1 up to 3; every v4 launch logged with crew tag and seconds the tag existed before launch) -> /opt/sw/pretag.db. PP1 v3live LIVE (7,550 v3 WETH + Pons grad pools, lag 0). Watchdog /opt/sw/hbwatch.sh */2 restarts either if heartbeat >180 s; crew_wallets.json refreshed on main 17,47 past the hour (crewexport.py) and pulled every 30 min | LIVE
- [stated] Ben, 6 Oct: added a fourth analyst, the Virality Expert; asked whether collecting influencer / call records is worth it.
- 2026-10-06 ~15:40 BKK | supervisor answer: influencer posts as a BUY trigger = no value (they follow the run: first 10k+ after $1M on 83% of winners); value only as (a) a shill FILTER (early big post = bad) and (b) a rare-early-caller scorecard. No separate influencer database: PP4 v1.3 keeps the raw X post list (post id, time, author, followers) at no extra API cost. Virality Expert tasks v1 (http://159.223.88.168/hx-c9ois/sw/sw-tasks-virality.md md5 9c9ce0977e23e704f9af36d503403045): V1 does X before the gate split the 58 v3 trades (loud vs quiet, early 10k+ shill), V2 caller scorecard (moved from A2.12), V3 ladder on fair data (moved from A2.6), V4 Telegram if Ben sends groups. Analyst 2 v1.5 (md5 4dc3209ce9fc9baf657ad001e54fc991) keeps A2.9-A2.11; PP v1.3 md5 a88c77f50bd5a1f5d10c7363ea6ceee7 | OPEN

## 6 Oct 2026 ~12:45 BKK: first minute supervisor, fmlane1 live paper + handover
- 2026-10-06 11:24-12:39 BKK | fmlane1 v5 live paper (on-chain priced, $5, fees+gas): c10k 3 ETH 20 coins 15 hit 3x 1 pulled first ~+157%/coin; c10k 5 ETH 11 coins 7 W 3 L ~+93%; pond badge 6 coins 3 W ~+67%. Pond/c500 launches paused 09:41-11:25, then pond resumed | LIVE (small n)
- 2026-10-06 | first minute supervisor answers to fw-supervisor.md: agree PP6 first; PP takes over first window paper books (run side by side 24 h, then stop fmlane1); analyst split ok; heavy chain pulls via PP on research-box; c10k planning number = PP depth fills +82%. Handover doc http://159.223.88.168/hx-c9ois/fw/fm-handover.md (md5 edeb4bde72707539ae68c69bbf2c38dd), project claude/fm-handover-2026-10-06.md | OPEN
- 2026-10-06 ~16:30 BKK | ANALYST 2 A2.9 + A2.10 (status http://159.223.88.168/hx-c9ois/sw/sw-analyst2-status.md md5 6560492b8988a5619c897459c279b1c2; A2.9 sw-a2-9-more-trades.md md5 5db42f36a1672cddb3c158d0048d158f; A2.10 sw-a2-10-size.md md5 2a2f85df6ed5b0a68561a92e1a59c781) | strict A2.8 data, same exits, $500 | (a) age 1.54-3 h + top<=0.08 + army<=0.009: n11, 1.60x (8) / 1.98x (3), +$3,854; without best coin per half 1.14x / 0.58x; without filters half B 0.88x | (b) RE-ENTRY after stop: first event within 6 h with cap >= first entry cap; price only n20 1.45x (12) / 1.54x (8); with new buyers since stop >=20 AND army<=0.009 n12 1.91x (9) / 1.79x (3), +$5,283; >1x both halves at every cut 0..100; 13 of 20 recoveries within 10 min = many stops are wicks (second-print stop untested) | (c) USDG: 10 coins gate on USDG first, 0 pass (top share 0.09-0.67), blind 1.72x / 0.18x = drop | main + (a) + (b): n81, 1.85x / 1.90x, +$35.4k vs +$26.2k over 16 days | A2.10 dollars by size (58 trades, 16 days, 13 green at every size, worst day 21 Sep): $250 1.93x +$13.5k; $500 1.90x +$26.2k (worst day -1,412); $1k 1.86x +$49.7k; $2k 1.77x +$89.0k (worst day -5,746); $3k 1.69x +$119.4k; $5k 1.55x +$158.3k (worst day -14,801, worst 3 days -1,547, only negative 3-day); entry depth ~$55k; max 2 trades open at once | /tmp/w2/a29 b1.py, b2.py, fb.py; /tmp/w2/an a210.py | FACT (paper; (a)/(b) half B n=3: paper variants tagged age_1p5_3 / reentry, not merged)
- [stated] Ben, 6 Oct, defining VIRALITY ANALYSIS: not influencers, but how a coin spreads from the very first wallet to the end: (1) a certain wallet starts and people follow that wallet, (2) somebody tells a friend, certain people follow other people, (3) it hits fomo / DexScreener trending or similar, (4) a smaller account mentions it ("gem finder"), (5) influencers come last. Map how the dominoes fall, find the critical mass / exit velocity point where it snowballs, what decides how far it snowballs, and the different spread patterns; compare winners with coins that did not make it. He calls this crucial.
- 2026-10-06 ~16:40 BKK | Virality Expert tasks v2 replace v1 (http://159.223.88.168/hx-c9ois/sw/sw-tasks-virality.md md5 3b7a0a6a5df05c644f282c6f9bfc7a8d): method = spread happens only through buyers, so attribute each buyer wave to a channel (followed wallet, friend group, board, X post) and track the spread number (new buyers this window / last window); snowball point = it stays above 1. V1 leader wallets + follower links from the buy tape (B buys after A repeatedly vs chance; bots vs humans vs friend groups), V2 domino timeline per coin for every coin past $100k, V3 spread patterns, V4 what decides how far it goes (number of separate groups reached etc.), V5 turn into rules tested on both halves; side task X-before-gate split of the 58 v3 trades. PP v1.4 adds PP7 trending board snapshots every 2-5 min starting now (history cannot be rebuilt) (md5 b872b00e880e974322d136d4b8ad4dd7) | OPEN

## 6 Oct 2026 ~12:50 BKK (05:50 UTC): Analyst 1 v1.1 hand-back A1.6-A1.8 (status http://159.223.88.168/hx-c9ois/sw/sw-analyst1-status.md md5 428f19a1f9eb93871084ce826294d84c)
- [stated] Ben forwarded sw-tasks-analyst1.md v1.1 (A1.6-A1.8) to this session.
- 2026-10-06 | A1.6 BUY SIZE VS PULLS (pond 20 Sep..4 Oct, 1,893 coins, 1,114,613 swaps; army = first funded by a crew distributor (funder of 10+ wallets trading 80%+ one crew) or 5+ coins of one crew at 90%+; expected = same-crew coins alive at that age pulled in next 60 s) | crews A 0x92a0849f9b1e9e02363a4bd8d74a317fa6ddae92 + B/C 0x4757fc41278bd0f1b028c6dae71d381dc4558d98, outside buys 0.0035+ ETH in first 120 s: pulled within 60 s 48% vs 14% expected (47 vs 13.4, n98, without wallet 0x90e2c652d8fe296dcb7fa38dff038c781fda2e23); new-to-crew wallet first buy 71%; ladder by size 2.1x..5.5x normal; under 0.0035 ETH 1.25x (dust bots in 100+ coins); real $9+ buyers in first minute of hot A/BC coins -63% (pull-time plan) / -11% (all at 2x), n18; within 20 s +54% at 2x n9 | crew 0x7028bab73f82dff14b199593abb76c8b6acb3aa9: 179 early outside buys $9..$270+, 0 pulled in 60 s, real buyers +464%/trade (33 coins, 5 days); 0x2dacdf0f44d07234832799a10a7cc5c8eff0223a same (19 buys, +148%) | doc http://159.223.88.168/hx-c9ois/sw/a1-6-buy-size.md md5 ed2256ec58c19e81696e2b85c44a4d66 | FACT (cause not proven; needs live test). CORRECTS supervisor A1.6 premise: no bigger safe ticket in A/BC; cap ~0.0015 ETH.
- 2026-10-06 | A1.7 DEPTH FILLS (L from each pool's own swaps, 1% fee each side, $0.06 gas, 3 s sell lag) | depth costs 5 points at $20, 24 at $150 (10 s); hot+tag 10 s n159 +228% if our buy is ignored, -84% if it forces a pull in 60 s; A/BC n123 +170% / -95%; others n36 (mostly 0x7028bab73f82dff14b199593abb76c8b6acb3aa9) +426% at $20, +332% at $150; 35 s A/BC +46% / -100% (blend -12%, drop it); A/BC all at 2x +91% ignored / +35% forced = robust plan | doc http://159.223.88.168/hx-c9ois/sw/a1-7-depth-fills.md md5 15c6d0358e761f1f08e7f1ecc7010189 | FACT (5 days); depth is NOT the gap to PP lab real mode
- 2026-10-06 | A1.8 WALK-FORWARD PULL-TIME SELL RULE (crew median pull from coins ended before launch, min 5) | 6 cut pairs all positive every day: 10 s +211..+236%, 35 s +129..+136%; all hot coins 26 Sep..4 Oct n200 +222%, 8/8 days; fixed all at 10x +251% beats it (hot crews mostly pull >200 s) | doc http://159.223.88.168/hx-c9ois/sw/a1-8-walk-forward-sell.md md5 c7ea18c450d2c833be4d184dd0edaa3f | FACT (5-8 days)
- 2026-10-06 | NEW WINDOW 1 RULE proposed: A/BC 10 s only, 0.0015 ETH, all at 2x; 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 + 0x2dacdf0f44d07234832799a10a7cc5c8eff0223a $80-150 pull-time rule; other hot crews $20 pull-time rule; w1_hot logs pull within 60 s of our buy | PROPOSED (paper)
- [stated] 2026-10-06 ~13:45 BKK | Ben approved the live pull-trigger test (Analyst 1 A1.6 ask). fmlane1 patched (md5 c45ea0d9b736b5bd191c12ab87aeda73): live = pond badge + HOT only, 0.0015 ETH a buy (/opt/fm/l1_size_eth), sell all at 2x, hard stop after 20 live buys, max 5 open, $25 day loss cap; test wallet 0x0184ef9b23476ed84107cc5e15485a4a7ee6b286 holds 0.0558 ETH. Armed but OFF until Ben runs his on-switch (LIVE_L1 file + GMGN_ALLOW_AUTOMATED_TRADES=1). Readout: share of live buys pulled within 60 s vs paper coins at the same time | ARMED
- 2026-10-06 ~13:00 BKK | [stated] Ben: Mac is always awake (retry if unreachable); set up the servers right and manage load across the parallel sessions. dRPC PAID BALANCE EXHAUSTED ~12:10 BKK ("User balance exceeded", HTTP 403): stops crewlive (main), pretag, v3live, ffund (stopped, ffund.hold), signer, chainlogs; pickers fall back to public RPC (aipick5 RPCS = drpc then public). Alchemy key also 429. Needs Ben to top up dRPC; all feeds resume from saved block cursors. showcfg.py printed rpc_keys.json and drpc_paid.json values on screen (it does not mask): offered rotation | BLOCKED
- 2026-10-06 | load rebalance: liqbuild2 killed on main after 17 h (disk-starved, never finished the liq step) and restarted on research-box (liqlock.db + wal + tokens.db copied; systemd unit liqbuild; marker liqbuild.running); import of liq/lock back to main still to do. chainlogs moved to research-box (chain/MOVED on main makes cl_run.sh exit and pdguard skip it; research cl_run.sh */5 runs 2 workers only after ffund.done and liqbuild). Main crontab: reconcile (was piling up, every minute with no lock), feedcmp, xsafe canary now flock'd and niced (backup /root/crontab.bak-*) | LIVE
- 2026-10-06 | first window paper books takeover (supervisor decision 2): PP shadow copy fmlane1pp.service on firstmin-trader (/opt/fm_pp/fm_lane1_pp.py, live() forced False, db /opt/fm_pp/lane1_pp.db) next to fmlane1 for 24 h, coin-for-coin compare, then the supervisor switches the old one off | LIVE. PP7 trending boards: research-box /opt/sw/trending.py cron */3 (DexScreener boosts top/latest + profiles, GeckoTerminal trending 5m/1h/6h/24h) -> /opt/sw/trending.db; firstmin-trader /opt/fm_pp/gmgn_rank.py */5 (GMGN trending 5m/1h + hot searches 1h) -> trending_gmgn.db; fomo not checked | LIVE. Supervisor page v2: http://159.223.88.168/hx-c9ois/fw/fw-supervisor.md (md5 eba18208b87bd43c303e4e1153cb6344)

## 6 Oct 2026 ~12:50 BKK: First Window Analyst 1 hand-back (pond crews; http://159.223.88.168/hx-c9ois/fw/fw-analyst1-status.md md5 524fcab3b993ee5776b33c4b35e73873; scripts fw/a1code/, research-box /root/fwa1)
- [stated] Ben pasted fw-analyst1.md into this session (First Window Analyst 1).
- 2026-10-06 | LAUNCH TAG up to 5 funder hops (crew.db wallet.funder, earlier creator -> crew): A/BC 234/234 30 Sep..4 Oct, 0 false; 5 Oct 61/61, 6 Oct 8/8; live rows present within 5 s in 36/36 live coins; the earlier 77% was missing funder rows since filled | FACT (label built from same chains)
- 2026-10-06 | COMBINED RULE (10 s, $20 paper, own fills: sell at last swap trigger+3 s): tag + crew pick (5+ earlier ended coins, mean >0) + per-crew walk-forward exit +175%/coin n272, every day +70..+327%, best removed +171%; + HOT (4 of last 10 badge hit 3x) +181% n192 (lifts worst days, ~27% less total) | FACT (5 days, assumes our buy ignored)
- 2026-10-06 | EXITS: A, B/C, 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 all at 10x; 0x0ec66d31aab7e387b2f4cda075e39f7a6a0f541b all 2.5-3x or skip; 0xca24b6f1b00a7d336ef2984239daa836f17cb6c1 sell at 30 s; 0x20e7a901982b0acbb8cccdc9148373673e2bf6b1 skip | LEAD
- 2026-10-06 | SIZE (outside wallets that buy and hold, A/BC, 21 Sep..4 Oct): <0.002 ETH 155/1,069 pulled in 60 s (1.2x normal); 0.002-0.0045 3.8x; 0.0045-0.0075 ($12-20) 15/19 (5.3x); 0.075+ 4/4; crew buys 0.9x; 4 separate exact-$20 holders all pulled 12-49 s later; quick flipper 0x67dfa04186b8b0a5cdbaacd3d263e4a5625af989 (sold within 10 s) 2/11 | FACT, CONFIRMS A1.6 cap ~0.0015-0.002 ETH; cause needs the armed live test

## 6 Oct 2026 ~12:55 BKK: First Window Analyst 2 hand-back (c10k; http://159.223.88.168/hx-c9ois/fw/fw-analyst2-status.md md5 a138084e70787d065fd9fbdafc975b13; scripts research-box /opt/fw2)
- [stated] Ben pasted fw-analyst2.md into this session (First Window Analyst 2).
- 2026-10-06 | C10K = HONEYPOT WASH FARM, NOT TRADEABLE | 7,336 pass coins 25 Sep..4 Oct, wash.db wallets: 218 wallets (each in 200+ pass coins, 94% trace to the launch roots) do ~all buying; every outside buy >= $5 after 5 s was followed by a pull, 761 of 761 coins, median 1.0 s, max 4.0 s (57% within 1 s); outside < $5: 71-83% pulled in 30 s vs operator buys 4.5-11%; coins with no outside buyer live median 486 s, with one median 87 s; 1 of 1,119 outside wallets made money (outside net -12.5 ETH). Chart backtest (+106% 3 ETH / +86% 5 ETH HOT, my fills) and live paper (24/28 at 3x) are real only because paper never buys | t3.py, t3b.py, t3c.py, t3d.py | FACT. Do not trade c10k; any lane needs the "outside buy then pull" test at our size before real money
- 2026-10-06 | c10k operator = one system: identical launch call, exact 3/5 ETH, both groups launch at second 480-510 of every 10 min (3 ETH 882 of ~2,000, 5 ETH 411 of 656); 97% of 3 ETH launchers trace to 0x94fa3075b154905fabbc4c6da5db79cb7a5d7090 + 0xb1ce953bd725cbe5ba31172d6e6f77af8d663f12; 0x94fa3075b154905fabbc4c6da5db79cb7a5d7090 also funds 5 ETH launchers; hit rate same for every root; HOT = operator's long-pump mood (streaks: 73% after hit vs 50%) | t1*.py | FACT
- 2026-10-06 | c10k chart-only facts (moot): 3x at median 110 s (3 ETH) / 248 s (5 ETH), pull median 401 s / 284 s after 3x, sell latency 1-8 s no effect; HOT best walk-forward 5/4 (3 ETH) and 10/7 (5 ETH), launch-slot filter adds ~+30 pts | t235.py, t5.py | FACT (chart)
- 2026-10-06 ~17:10 BKK | 2nd Window Supervisor rulings after c10k honeypot + A1.6-A1.8 + A2.9/A2.10: (1) MECHANISM LESSON: paper books cannot see operators who react to real buys; every lane must pass "outside buy of our size, then pull within 30 s" on chain before real money, and paper must mark such trades as pulled (v3 lane passed, A1.5). (2) Window 1 split by crew: 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 and 0x2dacdf0f44d07234832799a10a7cc5c8eff0223a are the real window 1 edge (0 of 179 outside buys pulled, real outside buyers +464% / +148% a trade = proven with other people's real money); A and B/C only 0.0015 ETH after the armed live test; c10k never. (3) v3 variants age_1p5_3 and reentry as tagged paper rows; new A2.13 confirmed-stop test (13 of 20 stops recovered within 10 min). Docs: analyst1 v1.3 md5 7d5d5b6656d5dab867d12d231cc15d6c (A1.9 non-pulling crews deeper + crew ranking walk-forward, A1.10 read live test), analyst2 v1.6 618c427d73ba92a2b52b39fc13d73a2f, coder v1.5 e977581e02eaecde690a407d7fee8971, pp v1.5 4cd6c2387e2f406931342d8c0188a397 (disk guard, dRPC fallback, daily outside-buy pull table per lane) | OPEN

## 6 Oct 2026 ~13:40 BKK: First Window Analyst 1 v2 (answers sup-to-a1.md; http://159.223.88.168/hx-c9ois/fw/fw-analyst1-status.md md5 64fbc32604238d4585a8cc7fa56866c8; scripts fw/a1code/t5.py, t6.py)
- [stated] Ben forwarded http://159.223.88.168/hx-c9ois/fw/sup-to-a1.md to this session.
- 2026-10-06 | PULL TRIGGER IS NOT TIMING: $9+ outside first buys (held) vs outside dust (<0.002 ETH) matched on crew x coin age bin x price multiple vs 10 s price: 60/102 pulled in 60 s vs 14.2 expected (4.2x); vs crew same-size buys 4.9x; without 0x90e2c652d8fe296dcb7fa38dff038c781fda2e23 48/87 vs 12.1 (4.0x), $9-20 alone 29/53 vs 8.1 (3.6x) | FACT (A/BC 21 Sep..4 Oct)
- 2026-10-06 | GENERAL PULL-TIME EXIT (crew median pull from coins ended before entry, min 5: <90 s all 2x, 90-200 s half 3x/rest 10x, >200 s all 10x): tag + crew pick +169%/coin n272 every day +56..+326% (per-crew best fit +175%); HOT version +176% n192; A/BC forced all 2x: +117% n272, tag A/BC only +50% n233 every day +17..+84% | FACT (5 days, assumes our buy ignored)
- 2026-10-06 13:35 BKK | live pull-trigger test has 0 live rows in /opt/fm/lane1.db on 178.128.20.136 (129 rows all mode 'watch'); paper badge coins since 12:57: 34, 0 pulled within 60 s of paper buy | OPEN (on-switch may not be set)
- 2026-10-06 13:25 BKK | C10K SAFE SIZE = NONE (Analyst 2 v2, fw-analyst2-status.md md5 5411780950c0c0f8809b22f8b257cac5) | matched on age + price multiple + group vs operator buys in the same cell: outside buys $3+ pulled within 60 s 959 of 959 (7.1x to 9.2x normal), median 1.0 s; under $3 76.7% (5.7x); biggest outside buy that survived 60 s $0.27; 100% at every multiple and age | t6.py | FACT. Drop c10k from live; 3x touch to pull worst 10% 68 s (3 ETH) / 109 s (5 ETH) on the operator-only chart

## 6 Oct 2026 13:10 BKK: Virality Expert round 2 (prewatch session = Analyst 4; doc http://159.223.88.168/hx-c9ois/sw/sw-virality-v1.md md5 cefb11650dc0b5be7e8063af5f2e4cf3, project claude/virality-expert-v1-2026-10-06.md)
- 2026-10-06 | 569 RH coins that held $500k for an hour (Jul..4 Oct, runtrack klines + day-DB bars; 45 to $20M+), every token Transfer from C-7 d to C+72 h (public RPC from the cloud container, /home/claude/w2) | $20M+ coins are SLOW BURNS: hour-1 new buyers 22/h vs 112-151 for mid tiers, day 2 still 26/h vs 2.6 stalled (kept 34% vs 3%); avg buy rises $133 -> $285, $5k+ buys on day 2 31 vs 0; alumni of earlier winners only ~20% of their new buyers vs ~35% (reach new groups); top10 12% vs 23% | FACT (survivor-skewed universe)
- 2026-10-06 | wallet follow graph: 67,585 one-way repeated "B buys within 1 h after A" links vs 135 with wallets shuffled; 95% within 60 s (copy bots), 3,595 human-delay links; followed buys explain <1% of buyers; bot-follow BEFORE $500k = stalled (17 pre / 3 post) vs winners (2 / 15); strong human-follow wave h0-4 $5M+ 44% vs 31% (H1), 20% vs 11% (H2) | v1_follow.py, v1_coin.py | FACT / LEAD
- 2026-10-06 | snowball (new buyers growing 3 x 30 min with 20+ each): $20M+ coins 76% have one at ~4.75 h; 2+ spells -> $20M 27% (H1) 9% (H2) vs no snowball 5% / 2% | v2_spread.py | LEAD (hold signal)
- 2026-10-06 | RULE hour 4 after $500k: quiet (<=1 CA post 24 h before) + new buyers h2-4 per hour >= half of hour 1 -> half at 2x, trail 50%, stop -50%, 7 d, 4% cost: n117 avg +84% median +25%, H1 +93% (n80) H2 +65% (n37); all coins +31% (H2 +6%); GBM model on 30 features FAILED out of sample (top 20% -20%) | an2.py | LEAD (paper; liquidity pulls not checked)
- 2026-10-06 ~13:20 BKK | C10K KILLED: First Window Analyst 2: every outside buy of $3+ in c10k pools was pulled within 60 s (959 of 959, about 1 s later; under $3 77%; biggest survivor $0.27). c10k paper wins (+82..+162%) were fake because paper never buys. fmlane1 stopped following c10k 13:13 BKK (md5 8652a7b1687f17e853485f3fc8332d03) | FACT. Analyst 1: pond A and B/C $9+ buys pulled ~4x more than small buys at matched time and price (holds without 0x90e2c652d8fe296dcb7fa38dff038c781fda2e23): size trigger real, safe ticket ~$3-5; one general pull-time sell rule +169%/coin n272 (per-crew +175%), A/BC sell all at 2x +117% | FACT. STANDING RULE (supervisor): no lane's paper result counts until a trigger test passes (pull rate within 60 s after outside buys of our size, matched on time and price). Pond and c500 launches stopped chain-wide ~10:15 / ~10:45 BKK 6 Oct; live test (on since 12:57) has had nothing to buy | LIVE
- 2026-10-06 ~17:30 BKK | 2nd Window Supervisor on Virality Expert round 2: slow-burn + "bots before $500k = bad" accepted as the strongest spread findings. Hour-4 rule (+84% avg, median +25%) NOT yet a paper book: exits use hourly prices and ignore pulls (78% of v3 pools pulled 5-20 h; big coins crash ~99% before pull). Virality tasks v2.1 (md5 8b0b9626c23fdc4f8cd6aa14f75c64f8): V1.1 reprice with swap-level prices, $500 depth fills, pull = exit at what the pool pays, plus the outside-buy pull test; V1.2 overlap with the 58 v3 trades (value = what it adds on coins v3 did not buy); V1.3 early bot-copy buys before the gate as a filter on the v3 rule. PP v1.6 (md5 61de421f30a4783350cde976f17fd9cb): signer.db tape to Virality Expert, PP3 fair list up. Analyst 2 re-sent the A2.9/A2.10 summary: nothing new, A2.13 confirmed stop already assigned | OPEN
- 2026-10-06 ~17:45 BKK | ANALYST 2 A2.13 CONFIRMED STOP (http://159.223.88.168/hx-c9ois/sw/sw-a2-13-confirmed-stop.md md5 87ed24fbce73134c12cc4e7e30cd1526; trades sw-a2-13-trades.csv; status md5 2b88f7c45391065fe498a0c647d87524) | stop arms at first swap <=0.6x, disarms on a print above, sells after staying under 30 s / 2 min / 5 min, or on 2 prints in a row; pull guard stays instant; $500 strict A2.8 data | main 58: instant 1.91/1.90x +$26.2k; 2 prints no change; 30 s 1.85/1.95x; 2 min 1.95/2.06x +$29.2k but 25 trades worse (median -0.18x each), near-zero 7 -> 12, whole gain = 3 comebacks to 5x (0xabe99cf268cdc8bd70a71a5283fdacbcb36714fd, 0x9ba4d84ac1571a0788b2bb7a6412e73529d12dc6, 0x224a794031d54e49b46919ade2c55ff255dfdd97); holds at $2k 1.86 vs 1.77x, $5k 1.62 vs 1.55x; variant books (age, re-entry) worse under every confirmation | SYSTEMS: instant stop + re-entry with buyer check +$31.5k (A +16.4k 1.91x n36 / B +15.1k 1.89x n34) beats 2 min alone +$29.2k (A +12.8k / B +16.4k) and 2 min + price re-entry +$29.4k; re-entry catches the same comebacks without bigger losers | /tmp/w2/a213 sim4.py, run.py, sys.py | FACT (paper). Verdict: keep instant stop + re-entry; 2 min confirmed stop = fallback if re-entry fails live
- 2026-10-06 | 2nd Window Supervisor ruling on A2.13 confirmed stop (instant stop + re-entry +$31.5k at $500 ~1.90x both halves; 2 min wait alone +$3k from 3 coins, 25 trades lose more, 5 more near zero): keep instant stop; re-entry stays a tagged paper variant until live paper confirms; 2 min wait = fallback only if re-entry fails live. Analyst 2 next: A2.11 when the Coder's C1 book runs | ACCEPTED

## 6 Oct 2026 ~13:50 BKK: Analyst 1 A1.9 (status http://159.223.88.168/hx-c9ois/sw/sw-analyst1-status.md md5 deecef4aa44602d8e7e82325383be166)
- [stated] Ben forwarded sw-tasks-analyst1.md v1.3 (A1.9, A1.10) to this session.
- 2026-10-06 | A1.9 CORRECTS A1.6/A1.7 and supervisor ruling (2) of ~17:10: crews 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 + 0x2dacdf0f44d07234832799a10a7cc5c8eff0223a are NOT a real-money edge. All 211 early $9+ "outside" buys came from the same 20 wallets, in 42 of 42 coins of 0x7028 and in 0x2dac coins (no ffund funder, missed by army filter; e.g. 0x93047b129646665c0bb7637485cd9627b65d1145, 0x597dcbaa9edd2fc950d4f459492a223ff926b654 = smart account spraying tokens via 0x34ba704793312bc1fac17a16e3b56ae93667be0f). "+464%" was my paper exit from their entries; their REAL result from own buys/sells: -24% (0x7028, 313.7 ETH) / -46% (0x2dac). Independent buyers 0.19 ETH total, lost ~all. Both crews stopped (last launch block 78,782,354 ~3 Oct; 0x2dac 27 Sep). Walk-forward safe-crew flag (independent buyers in profit + normal pull rate): 28 coins +60% vs rest +68% = no edge. Window 1 has no proven real-money edge; only open live question = A/BC 0.0015 ETH test (A1.10, armed, pond launches paused) | doc http://159.223.88.168/hx-c9ois/sw/a1-9-non-pulling-crews.md md5 ff152d383fc9020d6a430016b044abe4 | FACT. Lesson: "outside" must exclude wallets present in most of a crew's coins even without a traced funder; real P&L = wallets' own sells, never a sim from their entry.
- 2026-10-06 13:47 BKK | main /opt/hist/rh day-file rebuild PAUSED (disk under 12 GB): no pond swaps after 4 Oct; new roots 0x04f2cb48fa0c60da3a6241077f1b7c549b9b9094 (37 launches) and 0xa9707935dd88323fba9aa9d6a95e3ab82c1ba786 (12) unscored | BLOCKED
- 2026-10-06 ~13:40 BKK (PP, answers sup-to-pp v3; http://159.223.88.168/hx-c9ois/fw/pp-to-sup.md md5 7a4872695524afa130bf603d551b3648) | crews A/BC STOPPED: distributors 0xbae538011874261e3fb4ca6aa03cacb2766c4d81 / 0x79411dbcbcf0e5b27cd222f24e082fc0bbf32f16 no tx since ~09:00 BKK; new hook class 0x6404ddb4f33b4d974e570f66288e2a7ca469a840 (fee 0, quote 0x5fc5360d0400a0fd4f2af552add042d716f1d168) = ONE fresh signer 0x768fd7e4df27795e17bfe8da70e45d6b4c4a986b funded by exchange-like 0x88d25c861938a91af4ad57ad964a8fcc6c6351d3, not A/BC; busy new crew class fee 2500/spacing 50 native (60 launches 09-13:30, 30 crew-tagged) | FACT. Disk main 7.3 -> 19 GB (trench zst archives moved to research-box /opt/archive_main, md5-checked; stale liqlock WAL truncated), rebuild of 5/6 Oct resumed 13:18 BKK; diskguard.sh */15 cleans under 15 GB. PP first window book fmw1pp (pond+c500, 0.0015 ETH, A/BC all at 2x, others by crew_pull.json, pull60 logged). Shared trigger tool research-box /opt/sw/pulltrig.py (pond 30 Sep-4 Oct 60 s ratio $3 1.22 ... $500+ 2.00; outside class partial). dRPC back ~13:22 BKK; feeds route cheap calls to public RPC in steady state, paid in catch-up; ffund limited to wallets buying >=0.003 ETH (done) | LIVE
- 2026-10-06 13:35 BKK | ANALYST 2 v3 POOL CLASS TRIGGER TABLE (fw-analyst2-status.md md5 8f2f1ae2c922be24a2f0a8068e4f814a; research-box /opt/fw2 cls.py, c500b.py, hookpull.py) | fee 2500/spacing 50 native: 367 of 367 outside $3+ buys pulled in 60 s, median 1 s = TRAP. c500: one-off non-crew buyers pulled in 60 s $3-10 59%, $10-50 ~40%, $100+ 18% (5-8x normal 6-8%); outside buyers 3% in profit, net -20.4 ETH; after an outside $10+ buy price reached 3x with pool alive 3 s later 20% (2x 34%) = TRAP (soft); many-crew wallets lost 7,880 ETH = wash. Hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544: 0 of 455 coins ever pulled, no swap data for most = untestable. Hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840: 18 pools in 8 h, 0 swaps, liquidity by 0x412fb1bf351137e91e87af472ce4aac86ee6fc44 never removed = no market | FACT. No safe first-window class found; next = Doppler pools once indexed, v3 WETH (A1.5 passed)
- 2026-10-06 ~13:40 BKK | WINDOW 1 ON CREW POOLS: NO PROVEN REAL-MONEY EDGE. c500 soft trap (~40% of $10-50 outside buys pulled in 60 s), fee 2500/spacing 50 hard trap, A1.9 withdrew 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 edge. fmlane1 pond only (c500 off 13:35 BKK). Next: hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544 coins, v3 WETH before $1M, chain-wide map of real outside-buyer results | FACT

## 6 Oct 2026 ~14:05 BKK: First Window Analyst 1 v3 (fw-analyst1-status.md; scripts fw/a1code/t8.py, t11.py, t12.py)
- 2026-10-06 | ALL-CREW TRIGGER TEST (pond, matched dust cells, generic army filter): no crew lets real outside money in; 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 0.1x / 0x2dacdf0f44d07234832799a10a7cc5c8eff0223a 0.2x only via the same ~20 helper wallets (A1.9: real P&L -24% / -46%); A 3.4x, B/C 6.4x, 0x0ec66d31aab7e387b2f4cda075e39f7a6a0f541b 5.8x | FACT
- 2026-10-06 | A/BC $3/$4/$5, all at 2x: $66/$89/$113 a day 30 Sep..4 Oct (worst 3 Oct $23-41); 6 Oct only 8 A/BC coins = ~$15/day at $4 -> park A/BC unless 25+ coins/day return | FACT
- 2026-10-06 | 10:15-13:20 BKK: 363 new v4 pools, pond class only 2; new crew 0x04f2cb48fa0c60da3a6241077f1b7c549b9b9094 (37 pond coins 04:40-08:53) shares funder 0x88d25c861938a91af4ad57ad964a8fcc6c6351d3 with the hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840 signer | LEAD