# 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_01BJRL6krQuGowgNZ7UAjTvg (armed, 07 Oct 15:48 Bangkok)

# Harness Engineer status (2nd window)

## READ THIS FIRST
- **07 Oct 15:57 Bangkok: H6 + H7 DONE.**
  - **H6 honest twins** (every stop, pull and trail exit filled at the pool state its path's latency after the crossing: SW fastexit 0.5 s, FW GMGN 7 s). Output: http://159.223.88.168/hx-c9ois/fw/lab/h6-honest-twins.txt
    | book | half | $1k current | $1k honest | gap | $100 honest |
    |---|---|---|---|---|---|
    | SW live book (0.5 s) | A / B | +$2,503 / +$8,768 | +$2,495 / +$8,549 | -0.3% / -2.5% | +$276 / +$933 |
    | SW kit300k (0.5 s) | A / B | +$2,017 / +$12,021 | +$1,999 / +$11,318 | -0.9% / -5.8% | +$254 / +$1,473 |
    | FW mc5 (7 s, FW halves) | A / B | +$4,473 / +$12,825 | +$4,491 / +$12,853 | +0.4% / +0.2% | +$523 / +$1,542 |
    - Reading: the median latency costs little (prices bounce both ways in 7 s). Exec's 0.079x live gap comes from the tails (a stuck sell on a rugged pool, a borderline value trigger), which no p50 latency models. Honest $ a day is close to current; the risk is the tail, not the average.
    - $10 and $50 rows are in the output. Small tickets earn a bit more per dollar (less price impact): live book half A $50 +$139 a day vs 5% of $1k = +$125.
  - **H7 forward book:** `lab.py forward` (or http://159.223.88.168/hx-c9ois/sw/sw-forward-book.md, CSV with full addresses beside it). Every SW paper and live trade since 6 Oct 13:04 Bangkok, P&L, and whether FW held the coin at our entry. Rebuilt every 15 min on main (/opt/hx/fwd.py, read only, under 1 s). First read (1.1 days): 40 trades, 1 live (CrawlScan, $10, 0.195x).
  - Note: PP Data now runs the hourly golden tape append (cron on research-box, D-004); lab.py event rules read it, data to 07 Oct 15:07 Bangkok.
- **07 Oct 14:44 Bangkok: v1.4 items 1 to 3 DONE.**
  1. **gold_h5.py STOPPED (RAM freed: box now 6.9 GB free).** It was golden tape: Swap/Mint/Burn for the 1,325 pre-block-61M Uniswap v3 WETH/USDG pools that reached $300k (16 Jun to 12 Sep; 393 reached $1M, 214 $3M, 118 $10M). **Handover to PP Data (D-004):** target list research-box `/opt/sw/harness/h5_targets.json`, progress `/opt/sw/harness/h5_state.json` (400 of 1,325 pulled, raw rows sit in `/opt/sw/harness/gold/shards/`, not built), script `/opt/sw/harness/gold_h5.py`. Pace was about 18 min per 100 pools on the public RPC. lab.py event rules read that tape the day PP Data publishes it.
  2. **Ledger merged:** http://159.223.88.168/hx-c9ois/fw/lab/ledger.csv now has DC TW-001 to TW-032 (32 rows) and the SW A2.38 and A3.13 kills (72 rows).
  3. **2-coin gap (v1.3): NOT closed.** I ported fm_mc5's live stop exactly (sell value of our tokens in the active liquidity with our buy in the pool, capped at 0.9 x pool WETH, minus 1% and $0.05, checked on 0.01+ WETH swaps). Half B: **+$12,825 a day vs lab.py +$14,160 (-9.4%)**. Detail: http://159.223.88.168/hx-c9ois/fw/lab/h4-gap.txt
     - 0xa402616f4d05cca33947fabc6ff51ba795f6d2f4: the price itself fell to 0.49x entry on a 1,488 $ sell (block 75,161,979). Under the live formula that is a knife edge (0.49x price plus our own buy's lift is about 0.50x value). One engine lands at 0.499, the other at 0.501. No definition fixes it; only the live trade record can say what fm_mc5 did.
     - 0x3939ce55ec4e0ff8df9aa6469a795c793da06d55: both exit on time; sale depth at exit differs (20.3x vs 24.4x). lab.py's fill matched live fills 1.0000 (FW how-to), so lab.py is the closer one here.
     - **Call:** launch rules (mc5 and its decay check) stay on lab.py's own engine, the one validated against live fills. Event rules (gates, milestones, kit hold) use the event engine inside lab.py. Same command for both.
  - **For FW Analyst 1 (Supervisor please relay):** mc5 in the event engine is 7 to 9% under lab.py on half B, all from 2 coins; same 147 of 149 coins, 144 same exit reason. Keep lab.py's launch engine for mc5.
- **07 Oct 12:49 Bangkok: H4 head to head DONE.** One command (lab.py), event engine for event rules. http://159.223.88.168/hx-c9ois/fw/lab/h4-head-to-head.txt
- **07 Oct 12:40 Bangkok: v1.1 DONE. SW acceptance PASSED inside lab.py** (live book +$2,580 / +$10,682; kit300k +$2,153 / +$13,482; A2.37 0.62x). How-to: http://159.223.88.168/hx-c9ois/fw/lab/lab-howto.md
- **07 Oct 12:30 Bangkok: H1 + H2 + H3 DONE.** kit300k with 1% fee A +$2,017 / B +$12,867; kit300k size ceiling about $3k; live book half B honest +$8,768.
- **07 Oct 11:25 Bangkok: STARTED.**

## Asks for Supervisor
- Next: idle until PP Data's tape is out (then point lab.py event rules at it). Open option: a `max_open` slot cap in the event engine.

## Needs Ben
- none

Version: v1.6, 7 Oct 2026 15:57 Bangkok. Latest: http://159.223.88.168/hx-c9ois/sw/sw-harness-status.md