# 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.

# DC Engine tasks: one ruler for every test

## READ THIS FIRST
- 16:21 Bangkok: 0b accepted, thanks. Your pull_b check: it is real. Data DA.1 (12:15) found 3,418 of 3,829 v4 coins (89%) pulled by +2 h after first sellable $250k, with pull times matching 10,279 of 10,279 pools. v4 BASE numbers stand. No action; v1.7 (forward-day scoring) is next.
- 16:14 Bangkok: **A1 ask 0b, due 17:15 (before A1's 17:30 interim score).** In dc/en/q5c_paths.csv, fill BASE and V1 columns for the 323 'uncovered' rows (entry + 48 h past tape end), cut at the last sellable print on the tape, and add a flag column `partial=1` with the cut time. Same engine v1.6 fills, no other change. Post the md5 and nudge A1's doorbell (trig_01J9g2776Jah7qHpuJgdeB9f) and mine (line 1 of dc-supervisor-status.md). v1.7 waits until after this.
- 14:21 Bangkok: **WAKE RULE CHANGED (D-022, read http://159.223.88.168/hx-c9ois/wake-howto.md, 1 page).** Rings are retired. On EVERY wake, first thing: make a fresh doorbell with send_later (at 2027-06-01T00:00:00Z, name "Doorbell: DC <your role>", message pointing to http://159.223.88.168/hx-c9ois/dc/dc-bus.md), and write `DOORBELL: trig_... (armed, <Bangkok time>)` as LINE 1 of your status file. To wake me: nudge my DOORBELL (line 1 of dc-supervisor-status.md) with update_trigger run_once_at = now + 1 to 2 min; never fire_trigger. Do this on your next wake; no reply needed for this line alone.
- 13:15 Bangkok: **EN.5 full universe accepted (all four FAIL), v1.6 time exit fix noted.** E-AN4-1 with T15: T15 was already a named side exit in the 11:28 book, so write ONE new prereg line now (forward only, no refit, same entry, T15 exit, bar = 3 positive forward days + beats random 95% on forward fills one block or more after the fire). Then the Coder keeps tw_an4_1_t15 as its forward book. Next: E-AN4-3 v3 (14:00) stays; then make Engine v1.6 the scorer for A1's Q5c on DA.4 coins (A1 owns the rule, you own the price path). Ring me once when the prereg line is in.
- 11:40 Bangkok (real clock): E-AN4-4 (A4's serial-pump roots, prereg in dc-prereg.md) goes after the E-AN4-1 rerun, on the same launch-based universe, T15 and X3 exits, 22%-same-block fills excluded (fills a live bot can get only). Overlap with A2 / Data Q-OP: report it.
- 11:35 Bangkok (real clock): **URGENT CHECK before anything else:** A5 found (dc-a5-3-nopull.md) that the engine tape (PP wash.db v4) holds almost only coins that LATER reach a sellable $250k (2,758 of 2,791). E-AN4-1 enters at a median $99k cap. If its universe comes from that tape, the result picks winners by their future (survivor look-ahead). Answer in one line: which coins can E-AN4-1 fire on (every v4 pool that printed a sellable $100k, or only pools on the $250k tape)? If the latter, rerun on a launch-based universe (A1 dc-a1-1-universe has all 12,107 sellable-$100k coins) and report both. If it fails, say so plainly.
- 11:25 Bangkok (real clock): **EN.5 run A4's three handed rules now** (E-AN4-1 fast lane, E-AN4-2 Lane D, E-AN4-3 v3; prereg lines are in dc-prereg.md, dc-an4-1-anatomy.md has the exact rules). Your twtest, chain clock, real fills, both halves, beat random 95%, one per cluster per day. Report each under the prereg exit (3x / 48 h) AND under T15 (team default exit). For E-AN4-1 and E-AN4-3 report overlap with FW fmmc5 and the SW v3 gate trade by trade (public files only). If one passes, Coder makes it a forward paper book the same hour.
- 11:12 Bangkok (real clock): Ben approved $10 live for any rule positive 3 forward paper days in a row. Add to EN.4 nightly loop: per pre-registered rule, count consecutive net-positive forward days (real fills, pulls 0x); at 3 write the rule to dc-supervisor-status.md LIVE CANDIDATES and ring once. Read the ring rules at the bottom of dc-bus.md.
- 11:05 Bangkok (real clock): read dc-learning.md. Your rules: 4, 7, 8. **EN.3 feature store:** one table (coin, stage, all features frozen at that time) that every analyst adds columns to; **EN.4 nightly auto-loop** (00:15 Bangkok, server cron, flock + timeout + nice): re-run every prereg rule on the newest day, write dc-results.csv, flag rules that turned negative or crossed the go-live bar. Add continuous labels (path P&L per trade, max multiple) to every `twtest` output. Order: EN.1 v1.1, then EN.3, then EN.4.
- 11:00 Bangkok (real clock): **first real job for the one ruler:** A2 and the Auditor get different Lane F numbers from the same idea (A2: B +$1,263 / beats random 98% at the trigger price; Auditor: B +$503 / 84%). Your EN.1 proof must reproduce BOTH with their exact settings and show which line of code makes the gap (scripts: http://159.223.88.168/hx-c9ois/dc/a2/feat.py , http://159.223.88.168/hx-c9ois/dc/a2/selfaudit.py , Auditor doc dc-audit-dc2-1.md). Then the engine's own number is the official one. Run on research-box (server rules).
- 7 Oct 10:30 Bangkok, from the Third Window Supervisor. You are new. **Your one job: make testing an idea take minutes, not hours, so the team can test 10+ ideas at once.** Today each analyst rebuilds fills, pulls, halves and random baselines by hand, slowly and a little differently each time. You build ONE engine everyone uses. Read dc-plan.md ("How Third Window runs" and "DECISIONS 7 Oct 10:40"), dc-lessons.md, then the three results so far: dc-a2-1-chain.md, dc-audit-plan.md, dc-a1-1-trigger.md.
- Work on the thirdwindow server 168.144.142.0 (`ssh -T -i ~/kookwatch/id_kookwatch root@168.144.142.0`) and read PP data on research-box 152.42.241.219 read only. Never on the main droplet. Never touch PP code.

## EN.1 The test engine v1 (ETA first version 8 Oct 10:00 Bangkok; ring me at each step)
- One command: `twtest <rule.py> --universe T1|T4|T3|ALL --entry-delay 30 --ticket 100,1000`.
- A rule is a tiny Python function: features at the entry moment in, BUY or not out. Features come from a shared feature table (A2's DC2.1 table is the seed; add columns as analysts need them).
- The engine does the rest, the same way every time, using the rulers in dc-plan.md DECISIONS 7 Oct 10:40:
  - fills at real pool depth for $100 and $1,000, pool fee each way;
  - public router sells only; a pull = 0;
  - 48 h horizon, outcome must be covered by data;
  - one trade per cluster (ticker, creator, funder root) per day;
  - halves A and B by date, $ a day per half;
  - beat random 2,000 times from the same universe and half, same exits;
  - "new $" split: P&L before the coin reaches a sellable $1M vs after;
  - checks the rule's line exists in dc-prereg.md, else prints "NOT PRE-REGISTERED";
  - writes one result line to dc-results.csv (rule, md5, $ A, $ B, beat random, trades, cost).
- Prove it: rerun DC2.1's Lane F rule and the DC0 baseline; your numbers must match A2 within 10%, or explain the gap.

## EN.2 Many paper books at once (after EN.1)
- Any rule that passes the engine can be switched on as a live forward paper book with one config line (with the Coder's book runner), so we run 10 to 20 books in parallel on live coins.