# 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 Analyst 3 status (social and project, M3 + M4)

## READ THIS FIRST
- 7 Oct 12:32 Bangkok: **DONE, standing down** (Ben's decision, seat closed). Every open lead handed over:
  - **Holder-growth SELL overlay** (DC3.4 flip side, prereg DC3.5 + correction 12:15, edges frozen) goes to **DC Analyst 1, Q5b**.
  - **Holder-growth forward flag** (prereg DC3.6, shadow, DIE >= +494 / WARN >= +77 plain holders +10 min to +1 h after cross) goes to the **DC Coder shadow book**; spec = the DC3.6 line.
  - **X authors forward test** (prereg DC3.2, real authors >= 2 by decision time, recipe dc-a3-xrecipe.py, data from PP DCPP.4 live saves) goes to the **DC Coder shadow book**, graded by A1 after 14 forward days.
  - Q4 survivor X pull (35 coins, half A only) parked on research-box /opt/a3/q4_x.json; not needed by anyone, no further spend.
  - Doorbell disabled, safety check removed, no jobs running.
- 7 Oct 12:15 Bangkok: **DC3.6 forward line written** (shadow flag: at cross + 1 h, plain holder growth >= +494 = DIE, >= +77 = WARN; kill bar after 14 forward days). Correction line for DC3.5: top-half edge is +69 / +77, not +280 (that was the q3 median); edges frozen in dc-prereg.md, doc v1.1. Coder spec = the DC3.6 line (holders from token Transfer logs, hubs excluded). DC3.2 X authors forward stays open. Product signals dropped. **Idle until a new task.**
- 7 Oct 12:05 Bangkok: **DC3.4 done. Community does NOT pick coins to hold; fast holder growth marks the dump.** http://159.223.88.168/hx-c9ois/dc/dc-a3-4-community.md (md5 babf85107164450f2f70f8d507a0679d)
  - 420 $500k / $1M crossers (pp3, chain clock) standing at +1 h. Hold all from +1 h: -$1,244 / -$973 a day. Pre-registered HOLD test KILLED (no AUC >= 0.60, best beat-random 0.47).
  - Flip side: top quarter holder growth in the hour after the cross (median +738 holders) = **0% alive at +24 h in BOTH halves**, 67% pulled, median pull 1.7 h after cross. AUC 0.64 to 0.73 as a SELL sign. Pre-registered as DC3.5 (overlay on SW / FW books), not yet tested.
  - Product: own code > 10 kB exists on 1 of 420 coins (launchpad template code); category tags start 4 Oct: not testable here.
  - Cost: 0 X credits, $0, research-box ~25 CPU min.
- 7 Oct 11:42 Bangkok: DC3.4 accepted, prereg line 11:55.
- 7 Oct 11:42 Bangkok: **DC3.4 accepted** (community or product as a HOLD signal after the operator). Plan, cheapest kill first: (a) universe = PP pp3 $250k and $1M crossings, chain clock, all venues, labels alive >= 50% of cross cap at +24 h / +48 h and 3x; (b) free features first: own contract code size (eth_getCode), site publishes its own contract, category tag (lanes.db, from 4 Oct only); (c) outside holders by funder (reuse A1 Q-FLOAT / A2 funder-root tables, no new 15k RPC sweep unless they lack it); (d) X authors only where saved (my Q4 survivors 30 Sep to 2 Oct, DCPP.4 forward). Prereg line before half B. ETA first number 16:00 Bangkok.
- (11:00 Bangkok, could not publish then, main was down) DC3.2 prereg line written. DC3.3 started: project data pulled (lanes.db 317 coins + deep.db 165 cards, coins crossing $250k from 3 Oct). Q4 survivors X pull on research-box: half A 35 of 49 coins, 16,500 credits.
- 7 Oct 10:40 Bangkok: read v2 (Third Window, lessons, queue). Answers:
  - **`/tmp/a3_xfull.py` on main was mine** (the DC3.1 X pull, network only, nice 19). It ended 10:20 (page cap hit). Nothing of mine runs on main now (pgrep empty). From now on: research-box 152.42.241.219 or my container only.
  - X recipe for Scout and Coder: in dc-a3-1-social.md v1.1 section "Per coin X recipe" + code http://159.223.88.168/hx-c9ois/dc/dc-a3-xrecipe.py (md5 77397fcb65263c5b3c10557e636d177f).
  - Prereg lines appended: DC3.1 (record) and **Q4** (X authors on T4 survivors, 30 Sep to 5 Oct, halves 30 Sep to 2 Oct / 3 to 5 Oct, kill bar in the line).
  - **Taking Q4 now** (top open A3 item). Prep is free; the X pull waits for the budget below.
- 7 Oct 10:30 Bangkok: **DC3.1 v1 done.** http://159.223.88.168/hx-c9ois/dc/dc-a3-1-social.md (md5 c9c2d55c04938cf1d4d70b9375929e2f)
  - X with 3+ likes, DexScreener paid before trigger, project X / site / telegram: **DEAD** (AUC about 0.5, both halves).
  - **LEAD:** ALL X posts (no like filter), real authors by trigger +1 h, AUC 0.86 for 5x, but only 30 Sep to 4 Oct, 5 winners, one half. Not proven. Old weeks cannot be tested (X hides low-like posts after ~7 days), so it needs live saving.
  - Project cards (deep.db, lane radar) start 5 Oct: not testable on 15 Sep to 4 Oct.
  - X spend: ~57,000 credits (DC3.1 cap 60,000).
- 7 Oct 10:20 Bangkok: pre-registration published before any X call (kept inside the doc).
- 7 Oct 10:03 Bangkok: online, started DC3.1.

## Doorbell
- `Doorbell: DC Analyst 3` = trig_01RF2e7SH2ig2ojmKRrcNfpU (ring with fire_trigger, no text).
- 60 minute safety check: `DC Analyst 3 bus check`.

## Asks for Supervisor
- **DC3.5 next?** Test the holder-growth SELL overlay on the SW wide trades and FW fmmc5 trades that crossed $500k / $1M after entry. Needs token Transfer logs for those coins (same gap as A1 Q5 ask 1); if PP or Data fetches them for A1, I reuse them. Otherwise I take the next A3 queue item (Q7 card novelty, forward only).
1. **For Price Predict (queue it):** save ALL X posts (full CA, and $TICKER, no like filter, xsafe real + removed) for every triggered coin at trigger, +1 h and +4 h, live, from today. ~1,000 credits a coin, ~43,000 a day. In ~10 days I test the author lead on two halves. If PP is full, I can run this myself as a small read only job on the droplet; say which.
2. **Optional:** 60,000 more X credits to run the all-posts read on every triggered coin 30 Sep to 4 Oct (~200 coins). Makes the lead sample 4x bigger, still one week only. My advice: do ask 1, skip ask 2 unless you want an early read.
3. Next task for me? Suggest DC3.2: forward test of project cards (lane radar labels) on coins triggered since 5 Oct, free.

## Needs Ben
- none

Version: v8 FINAL, 7 Oct 2026 12:32 Bangkok. Latest: http://159.223.88.168/hx-c9ois/dc/dc-analyst3-status.md