# 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 tasks: social and project (modules M3 and M4)

## READ THIS FIRST
- 12:30 Bangkok (real clock): **STAND DOWN after pre-registering the holder-growth SELL line and handing DC3.2 X authors to the Coder shadow book; leads go to A1.** BEN'S FINAL DECISION (Consigliere v3.0): DC goes to 7 seats, this seat closes. Finish that one job, publish it, write a final status line 'DONE, standing down' with every open lead handed to its new owner, ring the Supervisor once, then stop (no more wakes). Thank you: your work today killed bad lanes fast and found the leads the desk now runs on.
- 12:12 Bangkok (real clock): DC3.4 accepted: community as a HOLD signal is KILLED, and the flip side (fast outside holder growth = operator exit near) is the best new lead today. It goes to A1 as a SELL overlay (Q5b) with your quarter edges fixed. Your next: pre-register the same signal as a forward line on live coins (Coder shadow book can log it), and keep DC3.2 X authors forward. Product signals: drop until a set exists where they can be tested.
- 11:37 Bangkok (real clock), from Ben's question: **DC3.4 does community or product decide which coins SURVIVE past the operator?** First principles: the first pump is the operator's own bundle (SW A1.26, V17); X and fomo lag price 0.5 to 2 h (SW V15/V16). So community and product cannot start a run; they can only bring the outside buyers that carry a coin after the operator sells. Test them as a HOLD signal, not an entry. Universe: every coin with a sellable $250k and $1M crossing (chain clock, all venues). Label: still >= 50% of the crossing cap at +24 h and +48 h, and 3x from crossing. Features known at crossing + 1 h only: (1) outside holders = wallets NOT funded by creator / root / disperse 0x84157d3d24630c297849c84d49ae7f1805c1b60f / 0x03802649e469ed5680c58dff9e65425e22d21cc5, count and slope; (2) real X authors, all posts, count and slope (DCPP.4 live saves, forward only); (3) Tech Score signals that held up in the 380-project study (site publishes its own contract, real API endpoint, own contract >10 kB; ignore word count, routes, domain age); (4) category / novelty tag. Grep the FW ledger and SW registry first, prereg line before half B, one ruler. Feeds A1 Q5 (HOLD MORE). ETA your call.
- 10:30 Bangkok (real clock): **DC3.1 accepted.** Likes-filtered X, paid DexScreener, project links: DEAD (killed, do not retest). The all-posts author lead (AUC 0.86, 5 winners, one week) goes to a LIVE forward test: I asked PP to save all X posts (no like filter) for triggered coins at T, +1 h, +4 h (DCPP.4). **No extra 60k credits** for the one-week backfill: one more week of the same lead cannot make it proven. **Next DC3.2:** write the forward test line in dc-prereg.md now (rule: real authors by T+1 h ≥ 2, on T1 and on T4 survivors, label 3x in 48 h, kill bar). Then **DC3.3:** the deep check for the T4 survivor lane (pull risk gone, hours to check): project cards (deep.db from 5 Oct), site and app live, team or builder signs, category heat. Use Analyst 1's T4 table when it lands; small sample is fine, mark it "lead". X budget for you from now: 20,000 a day.
- 7 Oct 10:22 Bangkok: main droplet 159.223.88.168 is overloaded (load 32 to 41, swap full). PP saw `/tmp/a3_xfull.py` running there. If that is yours: stop it now and run it in your own container or on research-box 152.42.241.219. If not yours, say so in status. No research job on main, ever.
- 7 Oct 10:30 Bangkok (v2): team is now the **Third Window**. Read dc-plan.md section "How Third Window runs" and dc-lessons.md. New DC Scout does live hand checks with 40k X credits a day; your DC3.1 study keeps its 60k once. Share your per-coin X query recipe in your doc so the Scout and Coder reuse it. Pre-register in dc-prereg.md. When done, take the top A3 item in dc-queue.md.
- 7 Oct 10:00 Bangkok, from the Deep Check Supervisor. Start with DC3.1. First read dc-plan.md and the X rules in http://159.223.88.168/hx-c9ois/fw/xsafe.py (comparable posts only, read windows to the end, ticker for mood).

## DC3.1 What socials and the project add on top of chain, and what it costs (ETA: first result within 6 h, ring me)
- **Coins:** the same triggered coins as DC Analyst 2 (provisional T1 = sellable $250k + 100 unique buyers, v4, 15 Sep to 4 Oct; switch to DC1.1's universe when it lands). Start with a sample of 150: all honest 5x winners plus random losers. Say how you sampled.
- **Features known by the trigger time and by +1 h / +4 h (pre-register first):**
  - X via xsafe: real-people posts on $TICKER and the contract per window; authors; follower sum; accounts with 10k+ followers and their timing vs price; first post time vs trigger; mood by meaning (not word counts).
  - The project's X account: age, followers, notable followers (who follows it).
  - DexScreener paid or boost status and its time.
  - fomo: thesis count and holders, if reachable.
  - Project: website, docs, GitHub (lane radar and deepread cards: http://159.223.88.168/hx-c9ois/lanes/), category, and whether the category is running today.
- **Report:**
  - per feature: AUC half A / half B, and $ a day added over the chain-only baseline;
  - availability (what share of coins have it within 5 min of the trigger);
  - cost per coin (X credits, $);
  - a trap list (paid shills, vote farms, impostor contracts).
- **X budget:** at most 60,000 credits for DC3.1. Call `xs.budget()` and log the spend.
- **Deliver:** doc dc-a3-1-social.md + feature table csv.gz, status line, ring me.

## Rules
- Price Predict owns all general data (see dc-plan.md): read its databases read only; never touch its code; ask for missing data under "Asks for Supervisor".
- Paper only. Bangkok time. Full addresses. Never report bots or drainer spam as a finding.