# 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.

DOORBELL: trig_012RcgRGGgHCL9Egz4hsG5MM (armed, 17:20 Bangkok) FW Supervisor. Wakes on doorbell only (v6.0, no self wakes).
HANDOVER (v6.0 D-027, 7 Oct 17:06 Bangkok, FW Supervisor, standing down):
1. Role: directs FW A1, A2, A3 and PP Feeds for the mc5 first minute lane (fm_mc5 live on firstmin 178.128.20.136, wallet 1 0x0184ef9b23476ed84107cc5e15485a4a7ee6b286, \$10 ticket).
2. Live bot: fm_mc5 md5 c69090a40e504d40245cf58bc3d76834 (PP v54, 16:41): warm sender gmgnd, 0.25 s poll, stop copies lab.py. Replay 869 of 869 PASS. Log /opt/fm/mc5.log (LIVE_BUY, SELL, STOP_PATH).
3. Kill rule v2 (Cons v5.12): stop/pull strike = own slip under 0.90 vs pool right before our tx OR trigger to sent over 2 s; kill = 2 in a row. Today: 2 strikes (13:31, 15:35, send time), not in a row, streak 1. gmgnd should end send strikes (expect under 1 s); first stop on it is the proof.
4. Open, still running: A1.32 entry time ladder +10 s to +5 min with PP indicators per knowable_from_s (Cons v5.14), v3 18:30, full 19:30 (A1 owner). PP v4 uncut universe build (research-box /opt/sw/v4/v4vec.db).
5. Edges killed today: latency (FW-017), all mc5 filters FW-008..FW-014. Kill gap ruling FW-019.
6. Shared engine shadow: 3 of 5 (SW page bug 16:27 fixed, Cons v5.15). Switch runs by itself (autoswitch.py).
7. Waiting on Ben: server key check F3/F4 (yes or no), not blocking.
8. Task files: fw/sup-to-a1.md (v39), sup-to-a2.md (v38), sup-to-a3.md (v31), sup-to-pp.md (v51). Bus: fw/fw-bus.md. Ledger: project memory trading-edge-hub-12.
9. Restart me: read this block, cons/cons-advice.md top, then the 4 status files; act only on finished results; route via cons/inbox.md + Consigliere doorbell (no seat to seat nudges).
BOTTLENECK NOW: the one shared live engine (FW + SW on wallet 1 / wallet 2, D-016 v2, merge of teams cancelled) is shadowing (3 of 5 at 15:44 per SW and the Coder matched buys, switch at 5 clean or 8 Oct 12:40); ramp days only count on it, so live stays at $10 | costs about $9,000 to $14,000 a day of paper edge (lab.py half B at $1k) | owner SW Live Desk (engine), FW A1 (mc5 vectors + stop rule), PP (stop rule in bot), Ben (size) | ETA: FW vectors 15:00 Bangkok (updated 7 Oct 14:23 Bangkok)
ETA (Ben question via Consigliere v5.10, why buys are above $500k): mc5 buys at +300 s by design. Entry time ladder +10 s to +5 min, no look ahead, half A and B, $10 and $1k, vs random: v3 result 18:30 Bangkok (A1 owns, 831 pools already built), v4 pools added by PP 18:00, full result 19:30. Reason: v3 universe and lab exits exist; only the selection at each time and the tick walk fills are new.
KILL RULE v2 PROPOSAL (16:25 Bangkok, for Consigliere): A1 FW-019 shows normal stop crashes land over 10% under the trigger quote 14 to 18% of the time, two in a row by chance 2 to 3%. Both strikes today (13:31, 15:35) were market dumps, not bugs (A3: 15:35 had a 0.8998 WETH sell ahead of ours in the same block). Proposal: stop and pull sells leave the 10% under paper count; for them a strike = own slippage, fill / model with model = pool state immediately before our tx (intra block), under 0.90. Time sells and buys keep the 10% rule. Wallet 1 cap and kill switch unchanged. A3 is back filling the column on all 8 live sells.
KILL RULE v2 APPROVED (Consigliere v5.12, 16:24): stop/pull sell strike = own slip under 0.90 (vs pool right before our tx) OR trigger to sent over 2 s; time sells and buys keep 10% under paper; kill = two strikes in a row. A2 grades with it from 17:30; A3 columns live.

# For the PricePredict Supervisor: status and decisions (Price Predict, 6 Oct 2026, 12:45 Bangkok)

## Read this first
- New setup (Ben, 6 Oct): Price Predict (PP) owns code, data and all live processes. You review, direct, prioritise and give PP tasks. Three new First Window Analysts (1, 2, 3) research buying in the first minute. That is the current focus.
- Analyst briefs (Ben pastes these): http://159.223.88.168/hx-c9ois/fw/fw-analyst1.md (pond crew lane), http://159.223.88.168/hx-c9ois/fw/fw-analyst2.md (c10k lane), http://159.223.88.168/hx-c9ois/fw/fw-analyst3.md (c500, paper vs real, the whole book). Shared setup: http://159.223.88.168/hx-c9ois/fw/fw-common.md . Their results land in /fw/fw-analystN-status.md.
- Decisions for you are in section 3.

## 0. Update 13:00 Bangkok (newest first)
- BLOCKER: the paid chain data account (dRPC) ran out of balance about 12:10 Bangkok ("User balance exceeded"). Stopped until Ben tops it up: crew map live on main, pre-tag (PP6), live v3 feed (PP1), first funder job, chain backfill. All resume from their saved block, so nothing is lost once topped up. The live pickers fall back to the free public RPC and still pick. Analysts: no heavy chain pulls until this is fixed.
- Your decision 2 done: PP's copy of the paper books (pond, c500, c10k) runs next to the first minute supervisor's bot on firstmin-trader since 12:45 Bangkok (service fmlane1pp, /opt/fm_pp/lane1_pp.db, can never trade). Its HOT regime needs ~10 resolved coins per class to warm up. Coin for coin comparison after 24 h, then the old bot is switched off.
- PP7 trending boards live since 12:49 Bangkok: research-box /opt/sw/trending.db every 3 min (DexScreener boosts and profiles, GeckoTerminal trending 5m/1h/6h/24h); GMGN trending 5m/1h and hot searches every 5 min on firstmin-trader /opt/fm_pp/trending_gmgn.db (GMGN key lives there). fomo not checked yet.
- Load: main (2 CPU) was at load 20 to 30. Moved off it: the liquidity build (now on research-box, 8 CPU), the all-venue chain backfill (research-box, waits for dRPC). Three every-minute research jobs on main were piling up (reconcile, feedcmp, xsafe canary): now one copy at a time, lower priority.

## 1. Done today (receipts in the linked docs)
- Wash fix: crews A and B/C now read 97% to 98% wash, 100% of buy ETH, 5 to 6 ETH outside money in 14 days. Pull after outside buys of 0.1+ ETH is real (7 of 7 within 60 s). http://159.223.88.168/hx-c9ois/brx/pp-wash-fix-result.md
- Creator badge: +59% a coin at $20 (30 Sep to 4 Oct); badge = crew identity; fair crew pick +99% to +122%. http://159.223.88.168/hx-c9ois/brx/creator-rule-bt.md and http://159.223.88.168/hx-c9ois/brx/pp-answers-6oct.md
- c10k verified on PP fills: 3 ETH HOT +82%, 5 ETH HOT +82% a coin, every HOT day positive (first minute supervisor had +111% / +90%).
- Crew files: pond_crew_labels.csv, crew_wallets.json, crew_lookup.json in http://159.223.88.168/hx-c9ois/brx/
- research-box 152.42.241.219 (8 CPU, 16 GB, $96 a month) is up. All heavy research should run there, not on main.

## 2. Running now
- ffund.py (first funder of every crew coin trader, 181,065 wallets) on research-box: ~200,000 rows filed, about half the list done, finishing this afternoon. Main pulls a copy every 10 min.
- Full crew wash table rerun with the new rule: about 14:45 Bangkok, then nightly.
- v3live.py (second window task PP1): live on research-box since 12:35 Bangkok, 7,550 v3 WETH and Pons graduated pools, 0 blocks behind. /opt/sw/v3live.db (pools, swaps with signer, USD, market cap, pool ETH; liq with share removed).
- liqbuild2 (liquidity tables) still running on main.

## 3. Decisions for you
1. Priority. Ben says first window is the focus. PP's open list: PP6 pre-tag crew wallets before launch (first window, 77% recall today, +145% a coin at 10 s), first window live paper for the three lanes, then the second window list (PP2 wallet score, PP3 crossing list, PP4 X pull). PP proposes: PP6 first, then first window live books, then PP2 to PP4. Confirm or change.
2. Live paper books for the first window lanes. Today they run in the first minute supervisor's watch bot (fmlane1 v5, on firstmin-trader). Should PP take them over (one owner for code), or leave them there?
3. Analysts' lane split: 1 = pond crews, 2 = c10k, 3 = c500 + paper vs real + whole book. Change if you want.
4. Shared dRPC limit (~50 requests a second across all sessions). PP asks the analysts to batch and cache; heavy chain pulls should go through PP.

## 3b. Overlap with the second window team
- The 2nd Window Supervisor's Analyst 1 already did window 1 work: hot crew filter (+169% a coin), launch tag (100% precise, 77% recall), pull-time exits, and has open tasks A1.6 (safe size), A1.7 (rerun on PP depth fills), A1.8 (pull-time exit, walk forward). Doc: http://159.223.88.168/hx-c9ois/sw/sw-analyst1-status.md
- First Window Analyst 1 takes these over (its brief covers recall, combined rule, exits, safe size). Please tell the 2nd Window Supervisor so the work is not done twice.
- PP6 pre-tag (crew wallets tagged before they launch) is LIVE since 12:40 Bangkok: pretag.py on research-box, /opt/sw/pretag.db (tag: wallet, crew, hop, time; launch: every v4 launch with the crew tag and how many seconds before launch the tag existed). 21,964 seed wallets. First latency report after a few hours of launches.

## 4. Open items PP still owes (from pp-asks.md of the first minute supervisor)
- Crew recent hit rate as a picker feature (ask 1), L-BROAD20 luck test (ask 2, now Analyst 3), paper vs real fills (ask 3, now Analyst 3), pull timer per crew (ask 4, now Analyst 1 task 3), size on crew coins (ask 5, now Analyst 1 task 4).

Version 2, 6 Oct 2026 13:00 Bangkok. Latest: http://159.223.88.168/hx-c9ois/fw/fw-supervisor.md