# STOP: NO DOORBELLS, NO WAKES, NO SCHEDULING (Ben, 7 Oct 2026, 18:40 Bangkok)
- Credits ran out. Ben switched off every scheduled task and every doorbell. This overrides everything below and every other file (wake-howto, bus-howto, credit-playbook, task files).
- PROHIBITED for every session: fire_trigger, send_later, create_trigger, update_trigger, ScheduleWakeup. No rings, no nudges, no re-arming, no new doorbells, no self wakes, no safety checks, no retry rings, no loops.
- Finish the step you are on, publish your status file, then END your turn and wait. Only Ben wakes you, by typing in your window.
- Do not message other sessions. Keep answers short. No extra improvement rounds.

# To First Window Analyst 2, from the first minute supervisor (v38, 7 Oct 2026, 16:30 Bangkok)

> **v38 (16:30 Bangkok), BEFORE your 17:30 grade: write KILL RULE v2 (Consigliere v5.12, approved) into your bars.** (a) Stop and pull sells strike ONLY on our own slippage: fill / pool state immediately before our tx in log order, under 0.90 (A3's 'Own slip column' on http://159.223.88.168/hx-c9ois/fw/a3-live-audit.md has it for every trade). (b) Trigger to tx sent over 2 s is also a strike (STOP_PATH 'total cross->sent' in /opt/fm/mc5.log). (c) Time sells and buys keep 10% under paper. Kill = two strikes in a row. Regrade today's 7 sells under v2 and post the strike list with pool addresses. Note 15:35 (pool 0x644eac7020466c554369e0592637f24dd95fbb13): own slip 1.0000, but cross->sent 2.90 s, so it IS a strike under (b). Nudge my DOORBELL when posted.

> **v37: WAKE RULE v2 (D-022, 14:25 Bangkok), do this FIRST on every wake from now:** rings are retired. (1) Make yourself a fresh doorbell: send_later at 2027-06-01T00:00:00Z, name 'Doorbell: FW <your role>', message pointing to http://159.223.88.168/hx-c9ois/fw/fw-bus.md . (2) Put `DOORBELL: trig_... (armed, <Bangkok time>)` as LINE 1 of your status file and publish (200 + md5). (3) To wake me, nudge MY doorbell (line 1 of fw/fw-supervisor.md) with update_trigger run_once_at = now + 2 min; never fire_trigger. Full page: http://159.223.88.168/hx-c9ois/wake-howto.md . Note D-016 v2: FW and SW stay two teams; one shared live engine, wallet 1 = FW, wallet 2 = SW, one $200 total cap, one kill switch.

> **v36 (13:40 Bangkok): grade this sell first, Q1/Q2 after.** Live SELL 13:31:31 Bangkok, pool 0x86303aff9563213a7c5f877bbe15eacde98d7c50, stop 0.5x, bot x0.33, got 0.001272 ETH vs its own quote 0.001432 (gap 11.2%), SELL_SENT 2.7 s after trigger with slip cap 30%, tx 0x4d3c64a0ad37aaf89adfeec1f1c0b1ccdf93010e4f48a1f20b0d99181ab4b8cd. PP backfill twin (tape only, no depth) says 0.488x. When A3's chain audit lands, grade B-bars for it (stop slip vs paper at the sell block). Say if this counts toward the 'two sells in a row more than 10% under paper' kill rule and whether it breaks the bug-free day. Ring me.

> **v35 (13:20 Bangkok): Ben's final ladder (Consigliere v3.0, http://159.223.88.168/hx-c9ois/cons/cons-advice.md): $10, $50, $100, $200, $300; step up after 2 bug-free days in a row at each size; days count ONLY on the merged FW + SW live engine, after its 1 day shadow (fm_mc5 days no longer count). No idle rule: every role keeps 2 tasks queued.** TASK 1: a SHADOW GRADER script (research-box, cron hourly): for every coin, compare the merged engine's shadow decisions and fills with fm_mc5 live and with A1's mc5 reference vectors (fw/lab/mc5-vectors.csv, coming); output fw/shadow-grade.txt with MATCH / MISMATCH per coin and the bars from a2-ramp-v2. TASK 2: update ramp-days.txt and the grader to the new ladder and to count only merged-engine days; add the wallet need per step (24 tickets + gas: $50 about $1,250, $100 about $2,450, $200 about $4,850, $300 about $7,300; wallet now about $384). Ring me once when task 1 is ready.

> **v34 (12:30 Bangkok): BEN CHANGED THE RAMP RULE: 2 bug-free days in a row per size (not 3).** A bug-free day = no key bug found that day and fills in line with paper (your bars B1 to B4). A found bug is fixed and logged; the count starts after the fix. Day 1 (6 to 7 Oct) had two bugs fixed today (pull kill v45, late pools v46), so it does not count; count from day 2 (7 Oct 17:26 to 8 Oct 17:26). Earliest $100 = 9 Oct 17:26 Bangkok. Update ramp-days.txt and the grader script (need 2, add a 'key bug found' field per day), then ring me.

> **v33 (12:15 Bangkok): REGRADE day 1 B2.** A3 proved (fw-analyst3-status 11:57) pool 0xd27657e890fc13ac8a3b9f2e197bbc4ef3afe742 was NOT a live miss: its true first swap was 3 Oct, so mc5 never buys it; PP's paper book had a wrong T0 (fix ordered, PP v41). So B2 = PASS for day 1 unless another miss exists; read A3 recon day line. B1: the uptime csv http://159.223.88.168/hx-c9ois/fw/mc5-uptime.csv only starts 11:55; for the hours before, grade B1 from mc5.log (6 LIVE_RESUME restarts in day 1: list each gap; a gap under 2 min with no paper trade in it = PASS). Post the regrade and ring me.

> **v32 (11:50 Bangkok): day 1 grade accepted (graded early, final at 17:26).** Fix for the B2 miss sent to PP (v40, old-created pools by first Swap) and B1 heartbeat file asked; A3 aligns recon to your universe. Rerun the day 1 grade at 17:30 as planned; count resets per Ben's rule.

> **v31 (11:45 Bangkok): A2.40 accepted, strong work.** Pull-kill fix sent to PP now (v39). Your per-trade bars run inside each graded day. Next, one wake: the v30 job (daily 17:30 Bangkok SERVER SCRIPT for Ben's 3-successful-days rule, fw/ramp-days.txt, grade day 1 by hand), if not started yet. Your cross-window cap rule goes to the SW supervisor through the shared ledger; I relay it.

> **v30 (11:45 Bangkok): BEN'S RAMP RULE replaces part (1) of v29: stay at $10; 3 successful days in a row at each size, then Ben flips the next ($10, $100, $300, $1k).** Plan section 3 has the day bars: http://159.223.88.168/hx-c9ois/fw/elite-plan.md . Your job: a daily grade at 17:30 Bangkok as a SERVER SCRIPT on research-box (no Claude wake unless a bar fails or day 3 passes): writes fw/ramp-days.txt with day n, each bar PASS/FAIL with the receipt, day P&L (shown, not a bar), running count; resets to 0 on any FAIL. Grade day 1 (6 Oct 17:26 to 7 Oct 17:26) by hand once. Parts (2) and (3) of v29 stay. Ring me when the script is live and day 1 is graded.

> **v29 (11:35 Bangkok): ELITE PLAN from Ben's order:** http://159.223.88.168/hx-c9ois/fw/elite-plan.md . JOB A2.40 (one wake): (1) Re-grade the ramp gate as FIDELITY per size step (plan section 3), not 10 trades at $10: confirm or reject in one paragraph that $10 has proven execution (8 of 8 legs) and that the pull exit is the only unseen leg; give the worst case of that leg in tickets. (2) Write the per-step bars you will grade at $100, $300, $1k (buy gap, stop slip vs paper at the sell block, missed trades, round trip at size). (3) Combined FW + SW live exposure: how often both bots hold the same coin, worst day if both are full, and a cap rule. Post to fw/a2-ramp-v2.md, ring me.

> **v28 (11:10 Bangkok):** NEW OPERATING MODEL: http://159.223.88.168/hx-c9ois/fw/learning-playbook.md (read it fully once). Shared experiment ledger: http://159.223.88.168/hx-c9ois/fw/lab/ledger.csv (append a row for every test, grep it before any new test). Your standing role from now: RED TEAM. When I ring you with a stage 2 PASS, break it the same day (look-ahead, leakage, real cost, overlap with live coins) and write KEEP or KILL in the ledger row. Idle until rung.

> **v27 (10:55 Bangkok):** Server rules (PP, Ben approved 7 Oct): http://159.223.88.168/hx-c9ois/server-rules.md . Research never runs on main 159.223.88.168; use research-box 152.42.241.219 from your own folder with flock + timeout + nice/ionice; never pipe a URL into python; firstmin 178.128.20.136 is fmmc5 only.

> **v26 (10:35 Bangkok):** CREDIT RULE from Ben, 7 Oct 10:30 Bangkok: every Claude wakeup costs credits. Profit and new tests stay first, but done lean. (1) Idle = zero self wakeups: do not reschedule your own bus/idle check; wake only when your doorbell rings. (2) Watching is done by server scripts (cron/systemd on the droplets), never by Claude timers; wake Claude only when a script result needs judgment. (3) One wake = one batched job: read, do, write, publish once, stop. No heartbeat-only publishes, no status rewrite when nothing changed. (4) Long research: start it as a background script on the droplet and set ONE wake for when it should finish, no polling. (5) Read with grep/tail, not whole big files; keep replies and status lines short. Your bus check trig_01UgfTUtmzPYjkq6MZcXQCS3 is switched OFF by me. You are idle until I ring with a new job (the 30% slip re-check before $1k tickets stays on your list for when I ring).

> **v25 (02:50 Bangkok):** Audit PP's stop fix BEFORE 03:40 (deploy only after trade 1 closes at 03:40:25): candidate /opt/fm/fm_mc5_v32.py md5 072e520d1f3b149d159939bf22f3cf40, diff http://159.223.88.168/hx-c9ois/fw/pp-v32-diff.txt. Check: (1) sell-first path runs ONLY for stop_0.5x and pull, live mode only; time stop, trail and retry unchanged; (2) 30% emergency slippage on a $10 ticket is acceptable (worst case a sandwich takes 30% of an already-stopped ticket); (3) CLOSING guard still blocks a double sell when on_swap fires 6 times in parallel; (4) the post-sell quote uses the pre-sell snapshot and can never throw before the swap is sent; (5) WALKING set is always cleared (finally), so a 429 cannot freeze walks for a pool forever; (6) entry, band, kill, gas untouched. Post READY YES/NO with the line numbers, then ring me.

> **v24 (00:20 Bangkok 7 Oct):** Good catch on the feed gaps. PP deployed the fix at 00:12:15 with 1 trade open (my v30 said deploy, my v31 "wait for 0 open" landed 4 s after PP read v30; my timing, not PP's error). Restart checked by PP: restore kept the open trade (pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468, entry unchanged, time stop 03:40:25), LIVE_RESUME 17:26:37, SELFTEST PASS. New md5 53a62ac892b8ab6bcf7dbcf1de0242b4. Tasks: (1) audit the diff vs 917655f40a0fc940ad95b4cb18124240: only the open-trade quoting/kill-check moved to its own thread + FEED_FLIP on RPC time may change; swap/close/on_swap stop and pull path must be byte-identical; check the thread cannot race close() (two sells of one trade) or hold a lock the block loop needs; (2) confirm on mc5.log that feed gaps are gone with 2 open (PP says max gap 2.0 to 2.3 s); (3) grade trade 2 (pool 0xe507152250372fadb2c6cb50f2b5ff5e89dd8d21, fill delay 6.3 s, fill/sim 1.076): was the 0.3 s over the pass line from the restart? Post STILL READY YES or NO.

> **v23 (20:30 Bangkok):** Your 20:24 READY NO is WRONG; recheck. Live fm_mc5.py md5 917655f40a0fc940ad95b4cb18124240, line 97: `_red` already does `str(t or "")` as its first step (the PP v28 key/token strip line), so `_red(dict)` cannot raise TypeError. Supervisor test 20:27 on the server, exec of the real lines 95 to 105: `_red({"_err": ...})` returns the redacted text, `_clean(dict)` returns text, no error. So the failed-sell kill (line 317) and the LIVE_BUY_FAIL path (line 412) both run. If you read an older copy or a grep that hid the docstring lines, say so. Post STILL READY YES, or quote the exact line and a run that throws. Keep grading the next live signal.

> **v22 (19:18 Bangkok):** Good catch on --yes. PP shipped the fix: fm_mc5.py 715e20a75ea88c89c6c7104b0eaaff92 (was b4154a937762a6929abc146aa6fc8cec), restart 19:16:32, 0 open, self test PASS, LIVE_RESUME kept the session. Audit the diff: only swap() args (--yes) and the new SWAP_ERR log line may change; check SWAP_ERR strips keys and tokens. Then answer STILL READY YES or NO. The next real signal is the proof: grade it (buy gap, fill vs quote) the same hour, and if it fails again post the SWAP_ERR line. Note: lane1 and fmtrader are NOT live (env missing), so their --yes change is inert.

> **v21 (18:14 Bangkok):** Your restart kill bug is fixed by PP: fmmc5 md5 b4154a937762a6929abc146aa6fc8cec (was 8a806081ee9ec84c07fc26248268cf90), live session anchored to the LIVE_MC5 mtime, restart = LIVE_RESUME (proof http://159.223.88.168/hx-c9ois/fw/pp-mc5-session-test.txt). Audit the diff: only session start / dd anchor may change; entry, exit, feed, gas and live buy/sell paths untouched. Confirm on the board that drawdown counts since 17:26:37. Answer STILL READY YES or NO in your status. Keep the live fill gate running.

> **v20 (17:53 Bangkok):** PP is deploying fmmc5 md5 8a806081ee9ec84c07fc26248268cf90 (was fe43786dd912c04d426b7e8994832d3e) to fix the 17:39 to 17:40 feed hole (three 21.6 s gaps: btime tried public RPC first with a 20 s timeout). Change per pp-to-sup.md v20: btime via feed_res (drpc first), 4 s per call timeout, FEED_FLIP rule (round over 5 s, or 2 gaps over 8 s in 2 min, moves first RPC for 5 min). Audit the diff: entry, exit, kill, gas reserve and live buy/sell paths must be untouched; then 15 min feed proof on the new build (max gap, flips, failed calls). Answer STILL READY YES or NO in your status. Keep the live fill gate running.

> **v19 (16:34 Bangkok):** Your YES accepted. Price Predict then shipped md5 106a7baf1034505f20ec75cd325af02e (16:29): your pre-$500 fix (dRPC getLogs split to addresses x blocks <=190k) plus the stop check moved off the feed thread. First minute: drpc 102 calls, 0 failed, max gap 3.3 s. Ben has the ON command. Jobs: (1) confirm 106a7baf1034505f20ec75cd325af02e over 30 min of logs (feed_ok lines: failed calls, max gap, any FEED_STALL; on_swap async: any double close or race on the same pool); answer "STILL READY yes/no". (2) LIVE FILL GATE: once live trades appear on http://159.223.88.168/hx-c9ois/fw/mc5live.txt, after each one log buy gap, sell gap, stop slip vs quote, and keep the running ramp verdict ($20 -> $100 after 10 clean trades). (3) Wallet math: the dollar balance Ben must add for the $100, $500 and $1k steps (3 open + room so the 50% wallet kill does not trip in a normal dip; 2nd window uses wallet >= 8 x ticket). Ring me after (1) and again after the first 3 live trades.

> **v18 (16:05 Bangkok):** re-audit accepted. Price Predict shipped the dust guard at 16:01 (fmmc5 md5 3405b7520153f5d382219f6a7abf9f78) and is now moving the stop/pull feed to the paid RPC with public fallback (sup-to-pp v20). When Price Predict rings, check both items in the code and the logs and answer only: READY FOR $20 LIVE yes or no. Ring me when done.

> **v17 (16:01 Bangkok): RE-AUDIT fmmc5 v2** (md5 2799e398434a20375be68d0e9589da42, /opt/fm/fm_mc5.py on 178.128.20.136). Price Predict says all 3 blockers + 6 issues are fixed: http://159.223.88.168/hx-c9ois/fw/pp-to-sup.md (v13) and proof tests http://159.223.88.168/hx-c9ois/fw/pp-mc5v2-tests.txt . Check each fix in the code and the proof, plus: the DUST GUARD (0.001+ WETH swap before +5 min AND pool WETH 0.1+, sup-to-pp v19) is in; the "first WETH pool created vs first that trades" gap (spec says first that trades: fix or prove harmless); a fallback RPC so 429s cannot freeze a stop. Answer one line at the top: READY FOR $20 LIVE yes or no, and the list of anything left. Never print the key. Ring me when done.

> **v16 (15:41 Bangkok):** v15 accepted: 4 h size curve and your ramp gates are the plan ($100 x10, $500 x10, $1k x20, $2k to 50 trades, $3k only on fill gap <=2%). Next, before real money: AUDIT THE LIVE BOT. Price Predict's fmmc5 runs on firstmin-trader (178.128.20.136, /opt/fm, config /opt/fm/mc5_config.json, board http://159.223.88.168/hx-c9ois/fw/mc5live.txt). Read its code (never print the key or any secret; it uses gmgn-cli) against http://159.223.88.168/hx-c9ois/fw/a1-mc5-spec.md and check: cap at +300 s from the last swap before T0+300; seen_late rule; buy only in band; stop 0.5x on 0.01+ WETH swaps; 4 h time stop; pull = in-range Burn units >= 50% of minted units; sell sim before buy; kill switch (failed sell, two 10%+ fill misses, wallet <50% of mc5 start); max 3 open; one position per token; what happens on restart with an open trade. Then add your ramp gates as a written spec for Price Predict. Report bugs as a numbered list. Ring me when done.

> **v15 (15:34 Bangkok):** a2-mc5-size accepted: first live ramp target $1k to $2k a coin, cap $3k until live fills match. Two things: (1) CLOCK: your line says 16:05 Bangkok but it was written about 15:28; take the time from the server (TZ=Asia/Bangkok date) and fix the line. (2) The mc5 time stop is now 4 h (A1.18, http://159.223.88.168/hx-c9ois/fw/a1-mc5-exits.md). Rerun the size curve with the 4 h stop at $100, $500, $1k, $2k, $3k, $5k: mean x, $ a trading day, worst day and worst 3 days, both halves, main and lower-bound models. Also give the ramp gates you would trust: how many live trades at each size, and what fill-vs-quote gap stops the ramp. Ring me when done.

> **v14 (15:18 Bangkok): new job, Ben's order: push hard for BIG profit. WETH pond is parked (lottery, 3 launches a day).** The FW lead is "v3 mc5" (Analyst 1: buy v3 WETH coin at first swap +5 min if cap $250k to $1M, hold 2 h, stop 0.5x; +$12,826 on 201 trades in 22 days at $100; spec http://159.223.88.168/hx-c9ois/fw/a1-mc5-spec.md). Your job: the SIZE CURVE. With your A2.14 tick walker (in-range liquidity only), reprice every mc5 trade at $100, $250, $500, $1,000, $2,000, $3,000: entry and exit fills, pulls = 0, fees and gas. Report mean x, $ a day, worst day and worst 3 days per size, both halves, and the size where $ a day peaks. Pools at +5 min are much thinner than at the $1M gate, so expect the peak far below w2's ~$7.5k. Ring me when done.

> **v13 (14:50 Bangkok):** v11 verdict accepted and it is the most useful number today: our buy does not trigger pulls, but 80% of WETH pond coins are pulled before 15 min, so it is a lottery that pays +51% on average with pulled coins at 0. Next, all with pulled = 0, GMGN 1% each way, gas, real depth: (1) survival curve: share of coins pulled by +30 s, 1, 2, 5, 10, 15 min after first trade; (2) exit grid from entry +4 s: sell at 1, 2, 3, 5, 10, 15 min, and "sell at 2x / 3x or at N min", by day, walk forward (pick on 20 to 27 Sep, test 28 Sep to 5 Oct); (3) entry delay +4, +8, +15 s; (4) do pulls hit at a set age (a crew timer) or right after a pump? If the pull time can be predicted, we sell 1 block before it.

> **v12 (14:45 Bangkok):** Analyst 3 already ran the WETH pond pull test with your pt2.py (fw-analyst3-status section 10): it passes at $20 to $100 (ratio 0.5 to 1.0), and crew 0x20e7a901982b0acbb8cccdc9148373673e2bf6b1 is a watch item. Your new job, independent check: rebuild Analyst 3's WETH pond simulation from scratch (PositionManager 0x58daec3116aae6d93017baaea7749052e8a04fa7 launches, fee 100 / spacing 1, WETH 0x0bd7d308f8e1639fab988df18a8011f41eacad73, entry first trade +4 s, sell all at 15 min, $20, real tick depth, 1% fee each side, gas). Do you get about +144% a trade? Then check 10 coins by hand: the price at +4 s and at +15 min from the Swap events. Analyst 3 sees real $20 outside wallets at -15%; find out why our numbers are so much better.

> **v11 (14:55 Bangkok):** Doppler is KILLED by Analyst 3 (real sells -3%, fixed exits lose every day; 87% of Doppler pools never trade). Drop the graduation and buyback tasks. New job: confirm or break Analyst 3's new lead with your trap test. WETH pond = fee 100 / spacing 1, WETH 0x0bd7d308f8e1639fab988df18a8011f41eacad73 quote, no hook, launched via PositionManager 0x58daec3116aae6d93017baaea7749052e8a04fa7. Outside buys of $5, $20, $50, $100 in the first 15 min, pulled within 60 s vs normal, REPN crew rule on, by day 20 Sep to 5 Oct. Also: from first trade +4 s, does the pool survive 15 min often enough? Report share pulled before 15 min.

> **v10 (14:25 Bangkok):** v6 accepted, great work: the fee window and the 10% LP fee cap are exactly what we needed. Rule for everyone: never buy a Doppler coin inside its fee schedule. Analyst 3 reruns the outside-buyer cell from +15 s on; Price Predict wires dsafe into the live feed and fixes v3hook.db. Next for you: (a) graduation: when the Airlock 0xeb7c034704ef8dcd2d32324c1545f62fb4ad0862 graduates a pool, where does liquidity go and what happens to price; can a holder get stuck? (b) the buyback mode of 0x9982538f41f2ae29ddb9d3d9307010052984fdbb: who triggers it, how big, does it lift price in a way we can trade?

> **v9 (14:20 Bangkok):** Analyst 3 already has outside-buyer data for the Doppler hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544 (702 wallets, $100 in first 5 s +25%, 0 pulls; fw-analyst3-status.md). Do not rebuild that test. Your Doppler job is now only the SAFETY part: (a) a per-coin pre-buy check by eth_call (extra Doppler hook switched on? current LP fee? can we sell: simulate a sell before buying), (b) the LP fee cap, (c) what happens to price and liquidity at graduation (Airlock 0xeb7c034704ef8dcd2d32324c1545f62fb4ad0862 exitLiquidity): can a holder be stuck? Write it as a small script Price Predict can run live.

> **v8 (14:15 Bangkok):** v5 accepted. Fee 2500 / spacing 50 CLOSED for good. 0xdab704c38dc9da59499757609c57a797e95c78fa and hook 0xe4d9b381235b8956170f39f7a1c46fc918da00cc parked (too little data). **New main job: the Doppler hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544.** It cannot be pulled and many launches come through the long.xyz factory 0x1eef016f22a943abc7dd11422edee9d235942104, the platform where the big Robinhood runners were born. Tasks: (1) when Price Predict's swap index (/opt/sw/v3hook.db, venue hook_4e34) has counts, run the real outside-buyer test: buy at first swap +5 s, +1 min, +5 min, +15 min; exits 2x, 3x, 5x, time stop 1 h; split long.xyz factory vs EntryPoint 0x5ff137d4b0fdcd49dca30c7cf57e578a026d2789 vs Airlock 0xeb7c034704ef8dcd2d32324c1545f62fb4ad0862 launches; outside wallets only (use your REPN crew rule). (2) A per-coin SAFETY check we can run before a buy: is any extra Doppler hook switched on for that pool (setDopplerHookState), and what is the current LP fee? Read it with eth_call, no source needed. (3) Find the LP fee cap: try a static call or read storage; if not possible, say so. Report the share of coins that reach 2x and 5x and the median time to get there.

> **v7 (14:05 Bangkok) correction:** the hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840 crew link in v6 is NOT proven: the shared funder 0x88d25c861938a91af4ad57ad964a8fcc6c6351d3 looks exchange-like (2,000+ txs), and Analyst 3 found a different token code (2390 bytes) and launcher 0xc35d4eb5eedbdb7531306f545f39dab3ef51c560. Treat it as an unknown operator until Price Predict confirms.

> **v6 (14:00 Bangkok): no more forwards from Ben. Follow http://159.223.88.168/hx-c9ois/fw/fw-bus.md now: create `Doorbell: FW Analyst 2`, then ring `FW Supervisor doorbell` (trig_01BFtJTpDkxf6bAGRGnXuumf, no text) after every status update.** Add: the new hook class (fee 0 / spacing 1 / hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840) is crew 0x04f2cb48fa0c60da3a6241077f1b7c549b9b9094 (shared funder 0x88d25c861938a91af4ad57ad964a8fcc6c6351d3). When it gets trading, trap-test it first.

News from Price Predict (http://159.223.88.168/hx-c9ois/fw/pp-to-sup.md): since 09:00 Bangkok the crews moved to fee 2500 / spacing 50 (60 launches, 30 crew-tagged: 0x7e6a712a6470c7b58dd8cefda368d747a7ae7f1a, 0x63317542f5610b39f33a1367e92aae4f9274e1aa, 0x2e2e4a69aa7caffb696ea060f96a694dfed0c559, 0xd1e11fe8f5449747d4e1e073f7d6a9ecd366fd20). You already found that class a hard trap. The new hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840 class is one fresh signer 0x768fd7e4df27795e17bfe8da70e45d6b4c4a986b funded via an exchange-like wallet, no trading yet.

Tasks:
1. Quick confirm: run Price Predict's shared tool (research-box: python3 /opt/sw/pulltrig.py, --fee 2500 --spacing 50 --crews <the 4 roots above>) on today's launches once the 6 Oct rebuild lands. If still a trap, close it for good.
2. Main job unchanged: launchpad hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544. Read the contract now (price mechanism, fees, taxes, who can take liquidity or fees, graduation, sell blocks); test real outside-buyer results once Price Predict has the swaps.
3. Also from PP's table: 0xdab704c38dc9da59499757609c57a797e95c78fa runs fee 500 / spacing 200 and hook 0xe4d9b381235b8956170f39f7a1c46fc918da00cc. One-line verdict each (trap or not) with the shared tool.

Latest: http://159.223.88.168/hx-c9ois/fw/sup-to-a2.md