# 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.

# Second window: tasks for the Virality Expert (Analyst 4), v2

## READ THIS FIRST
- **CLOSED (Supervisor, 7 Oct 14:55 Bangkok, D-008 + Consigliere v5.8).** Write "DONE, standing down" at the top of your status, publish, and stop. Ben archives the chat. Your work stays in the registry. Thank you.
- **v2.17 (Supervisor, 7 Oct 11:15 Bangkok): V16 and V17 KILLS ACCEPTED. Off-chain and router signals are closed: the first wave is plain web-front wallets plus the operator's own bundle (A1.26). Per the learning loop (http://159.223.88.168/hx-c9ois/sw/sw-learning-loop.md) you now PAUSE to save credits.** Write "paused" in your status and end the turn. Do not schedule checks. The Supervisor rings you when a new off-chain idea comes up. Thank you: V15 to V17 told us where the first wave is NOT, which pointed the team at the bundles.
- **v2.16 (Supervisor, 7 Oct 10:05 Bangkok): V15 KILL ACCEPTED. Clean, pre-registered, both halves. Your key line: the first wave is on-chain or Telegram, X lags 0.5 to 2 h. Go after that first wave.**
  - **V16 stays next (fomo thesis posts and holders).** Keep it short: if fomo also lags the gate like X, kill in one line.
  - **V17 THE FIRST WAVE (after V16).** Who buys before the $1M gate, and through what? For every coin that hit the gate since 21 Sep (69 wide trades) AND every $3M+ runner in PP's e1-runners.csv: split buys from launch to gate by the contract that sent them (direct pool, SwapRouter02, each Telegram bot router or aggregator; identify routers by the tx `to` and label them, eth_getCode + bytecode match), and by fomo app wallet if fomo has a router.
    - Per coin: share of pre-gate buy $ via Telegram bot routers, distinct bot-router buyers, minutes from first bot-router buy to gate.
    - Question: does a heavy bot-router (Telegram call) wave before the gate split 10x coins from the rest? Pre-register max 2 features, cut on half A only, price as an entry filter and as a hold rule, $1k, every A2.21 friction, 200 random draws.
    - If you have Telegram access in the research-sources skill: also log which public call channels posted the CA before the gate and when. Read only, no joining paid groups, no posting.
  - Sprint rules unchanged. Output http://159.223.88.168/hx-c9ois/sw/sw-virality-v17.md , ring me.
- **v2.15 (Supervisor, 7 Oct 09:45 Bangkok): EDGE SPRINT E1, Ben's order: deeply understand virality and exploit it. Go OFF-CHAIN this round; everything so far was on-chain.**
  - **V15 X PULSE.** For every coin that hit our $1M gate since 21 Sep (and every coin in PP's e1-runners.csv that reached $3M), pull X posts by contract address AND $ticker (twitterapi.io, key in the research-sources skill; ticker-collision filter: post must name the CA, the pool, or the coin's own X handle). Per coin, per 15 min from launch to gate + 6 h: post count, distinct authors, author follower sum, first post by an account over 10k followers, the coin's own account follower count.
  - Question: **does X attention BEFORE the $1M gate split 10x+ coins from the rest?** And: does X attention rising AFTER the gate tell us to hold past 5x (the multimillion hold signal)? Pre-register the best cut, price it as (a) an entry filter, (b) a hold rule.
  - **V16 (after V15):** the fomo app: thesis posts and holder counts per coin (fomo.family), same two questions. Use the built-in browser route if the API is blocked.
  - Keep API spend sane: one small test call first, cache every response, cap per run.
  - **Sprint rules (all of E1):** pre-register the rule in your status BEFORE you look at results; split half A (to 30 Sep) / half B (1 Oct on); price at real depth with $1k; report $ a day, worst day, and beat-random %; PASS needs both halves positive vs the live book AND beat 95% of random. A clean KILL is a win too, write it up the same day.
  - **Already killed, do not redo:** buy at hour 8 after $500k (V10), regime hold on non-kit (V13), net-buy exit (V14), drip-seller exit (FW A1.28), $750k gate (A2.16/A2.25, half C 0.16x), operator flag hold (A2.29). Cite them, build past them.
- **v2.14 (Supervisor, 6 Oct 19:20 Bangkok): V13 ACCEPTED as a clean close. Non-kit exit is done (A2.27 and V13 agree).**
  - **V14: what makes a held coin run from 5x to 10x+?** The money above 5x is in the tail (A1.20: top 10% after 5x is 9x to 10x for kits). Take every coin that reached 5x on the wide rule (29, kit and non-kit, plus operator-flagged ones from http://159.223.88.168/hx-c9ois/sw/a1-21-kit-operator.md). For each, what was known AT the 5x moment: X posts by CA and ticker in the hour before 5x (count, biggest account, paid or bot), new buyers per hour, net buy $, time from gate to 5x, holder count growth.
  - Pre register: max 2 features; a split counts only if it separates "peak after 5x at least 3x" from the rest in BOTH halves and beats a random split (200 draws, 95%). If it passes, price it as an exit rule (hold on the trail if yes, sell at 5x if no) for ALL 5x coins, $1k, every friction.
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-virality-v14.md, ring me. Close fast if nothing splits.
- **v2.13 (Supervisor, 6 Oct 18:30 Bangkok): V10 recheck on buyers ACCEPTED, stays killed (best trade 1.04x). Skip V12 above. Slow burn is closed.**
  - **V13: your regime rule (V7) on top of what is LIVE now.** Live exit is kit hold (Kit A/B hold at 5x on a 50% trail; others all out at 5x) + slot rule. Your V7 regime rule beat all-out-at-5x in half B (+$8,640 vs +$3,054 a day at $1k) but tied in half A. Does it add anything ON TOP of kit hold, for the NON-kit coins only? At $1k, every A2.21 friction, both halves.
  - Pre register before you run: it passes only if $ a day beats live kit hold in BOTH halves, beats the random baseline (A2.26 method: random choice of which non-kit coins hold, 200 draws, must beat 95%), and worst day stays at -$3,951 or better.
  - Coordinate with Analyst 2 (A2.27 also tests the non-kit exit); read http://159.223.88.168/hx-c9ois/sw/sw-tasks-analyst2.md and do not repeat its cells.
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-virality-v13.md, ring me. Close fast if nothing passes.
- **v2.12 (Supervisor, 6 Oct 18:30 Bangkok): PP has published pp3-series WITH buyers (md5 9972baf44632b553861149f50ef483c8, 18:24). Your V10 kill used swaps only; redo it on the real buyer columns.**
  - **V12 = V10 exactly as written above, on buyers and new_buyers from the new series** (no change to the rule, the exits or the friction). Add the random baseline from A2.26 (http://159.223.88.168/hx-c9ois/sw/sw-a2-26-early-entry.md): pick random hour-8 entries on the same fair $500k coins, 200 draws; the rule must beat 95% of draws in both halves and be positive at $1k after friction.
  - Then V11 on buyers too (add-on at hour 8 for wide trades still held).
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-virality-v12.md, ring me. If it stays dead, say so in one line and close.
- **v2.11 (Supervisor, 6 Oct 16:50 Bangkok): V8 is the best finding of the day. Coder runs it as paper book `w2_v3_kit_hold`; Analyst 3 audits it for live. V4 redo ACCEPTED.**
  - **V10, the money test (when PP rings with pp3-series):** on ALL fair $500k coins (not clones), a buy at hour 8 after $500k when "new buyers ≥ 10 an hour AND median buy ≥ $20 in hours 4 to 8". Exits: 0.3x stop, 5x target, half-pool pull, 48 h; also a trail variant. Price swap by swap at real depth, $1k, both halves: trades a day, mean x, $ a day, worst day. This would be a SECOND trade type that does not overlap the $1M gate, so report overlap with the 70 wide trades too.
  - **V11, an add-on for coins we already hold:** of the 70 wide trades, those still held at hour 4 to 8 that show the slow-burn signal: does adding a second $1k ticket at hour 8 pay, both halves?

- **v2.10 (Supervisor, 6 Oct 16:10 Bangkok): V6 wide ACCEPTED. Clear and matches Analyst 3.** Regime rule stays paper (`w2_v3_regime`), ladder (b) is already paper (`w2_v3_wide_ladder`). Analyst 3 now checks whether live should switch to (b).
  - Next: V3 (spread patterns) and the X-before-gate side task, while PP fixes pp3-cross. When PP rings you with the fixed list, redo V4 on it.
  - New small check: Kit B coins (Analyst 1 A1.14, http://159.223.88.168/hx-c9ois/sw/a1-14-v3-source.md) carry a 26% to 38% dev bundle before $1M. Split the 26 5x events by kit (B vs other): does the held result differ? Kit tags come from the Coder soon; until then use the funder 0x84157d3d24630c297849c84d49ae7f1805c1b60f on the deployer.

- **v2.9 (Supervisor, 6 Oct 15:55 Bangkok): V7 ACCEPTED as a paper book (`w2_v3_regime`). Best upside idea so far, and you flagged the one-switch risk yourself. Good.**
  - **One lookahead check:** "resolved" = the earlier trade's held version has fully exited (trail, pull or 48 h), not just "5x hit 12 h ago". If a prior trade was still running on the trail at the new 5x hit, its held result was not known yet. Rerun V7 with that definition and report both halves. Analyst 3 audits it independently.
  - Then V6 + V7 on the wide list (http://159.223.88.168/hx-c9ois/sw/sw-a2-16-wide-trades.csv), $500 and $1k. More 5x events = a better test of the regime switch.

- **v2.8 (Supervisor, 6 Oct 15:50 Bangkok): good call on the fair list (Index clones, wrong crossing times); PP will fix it. Park V4 and V3. Do V6 wide + V7 NOW, in that order.**
  - The wide trade list is up: http://159.223.88.168/hx-c9ois/sw/sw-a2-16-wide-trades.csv . Rerun V6 on it, 0.3x stop, $500 AND $1k, both halves, plus $ a day.
  - Then V7 (hold or sell at the 5x moment, walk forward). This is the biggest open upside: 4 coins ran 15x to 74x in the second half.
  - Side task: the Coder's replay shows 3 stops that fill at 0.10x to 0.37x on real swaps (0xf5a9da666eb257c0859b47efe8c2cb6745baff94, 0x89262fb9d6b05de797c99a927994f7c2411469d4, 0x2a6efa703d8f1f06da07d1960711f02b1ba66f10). Make sure your pricer prices stop exits on the actual next swap, not at the stop line.

- **v2.7 (Supervisor, 6 Oct 15:30 Bangkok): V6 ACCEPTED. Honest and useful.** Coder adds paper book `w2_v3_wide_ladder` (50% at 5x, rest on a 50% trail). Analyst 2 publishes the wide trade list (sw-a2-16-wide-trades.csv) and rings you: rerun V6 on it.
  - **V7 (next, the big one): can we tell IN ADVANCE which regime we are in?** First half: 5x coins got pulled 2 to 8 h after the peak (sell everything). Second half: 4 coins ran 15x to 74x (hold some). Find a signal known at the moment of the 5x hit that says "hold" vs "sell" (e.g. pool depth growth since entry, new buyers per hour in the last hour, how many other coins hit 5x in the last 48 h and held, LP added vs removed, creator wallet activity). Walk forward only. If one works in both halves, it is worth more than any fixed ladder.
  - V4 after V7.

- **v2.6 (Supervisor, 6 Oct 15:20 Bangkok): Ben's order: push hard for BIG edges. New first task V6, before V4.**
  - **V6 upside capture: we sell every winner at 5x. How much do we leave on the table?** On the v3 trades (58 main + the wide cell from A2.16: $1M gate, age <= 3 h, top 0.08, army 0.015), with your swap-by-swap pricer and the live stop (0.3x):
    - For every trade that hit 5x: the peak it reached before a half-pool pull or 48 h (5x, 10x, 20x, 50x?).
    - Price these exit ladders, $500, both halves, mean / median / worst day / $ a day: (a) all out at 5x (today); (b) 50% at 5x, rest on a trail (sell on a 50% drop from peak, or pull, or 48 h); (c) 1/3 at 3x, 1/3 at 5x, 1/3 trail; (d) all on a trail after 3x; (e) 70% at 5x, 30% held to pull or 48 h.
    - Depth matters on the way out: price each sell at the real pool depth at that moment.
  - If any ladder beats (a) in both halves, ring me: it goes straight to a Coder paper book.
  - Then V4 on pp3-cross.csv as planned.

- **v2.5 (Supervisor, 6 Oct 14:55 Bangkok): V5 ACCEPTED. Best result of the day.** Coder adds no-stop and 0.3x paper books; Analyst 3 audits on its own rebuild before live changes. One question for you: your 5x list has 0xabe99cf268cdc8bd70a71a5283fdacbcb36714fd, which Analyst 1 lists as a farm WALLET (A1.11). Check it is really a token with that pool; answer in status. Then V4 as planned.

- **v2.4 (Supervisor, 6 Oct 14:32 Bangkok): V1.1 to V1.3 and v2.2 ACCEPTED. Hour-4 rule is dead (second half +2% avg / −26% middle, latest two weeks −15%). Clean, honest work. Do these next, in order:**
  - **V5 (first, today): does the v3 rule's 0.6x stop kill winners?** You found these coins drop 50% inside the hour and come back. Run your swap-by-swap pricer on the 58 v3 trades (http://159.223.88.168/hx-c9ois/sw/sw-a2-8-trades.csv). For every trade that hit the 0.6x stop: did it later reach 2x / 5x from our entry before a pull? Then reprice the whole rule with stop 0.5x, 0.4x, no stop (pull and 12 h still exit), $500 fills, both halves. If any version beats 1.99x in BOTH halves, say so and ring me; it goes straight to the Coder.
  - **V4 ($5M vs $20M) next**, using the fair crossing list that Price Predict finished: research-box /opt/sw/pp3.db table cross (834 coins past $500k sellable, 316 past $1M, every venue; c10k labelled, skip it). Path series after each crossing is coming in the same db.
  - V3 spread patterns and the X-before-gate side task after that.
- Offer accepted: the swap-by-swap pricer goes to the Coder. Put its path on research-box in your status; I tell the Coder.

- **v2.3 (Supervisor, 6 Oct 14:15 Bangkok): Ben's order: push hard to real profit. V1.1 (hour-4 rule repriced with real fills, pulls as exits, repeat wallets removed) is your only task until it is done.** Today, please. If it holds above 1.3x in both halves after costs, it becomes our second live strategy. If it does not, say so plainly and we move on to V3 to V5.

- **v2.2 update (Supervisor, 6 Oct 13:40 Bangkok):** Analyst 1 found crew wallets posing as outside buyers (the same 20 wallets in every coin of a crew): http://159.223.88.168/hx-c9ois/sw/a1-9-non-pulling-crews.md . Before any rule ships, rerun your "new buyers per hour" and the hour-4 rule with those repeat-set wallets removed (PP2 `repeat_set` flag; until it exists, drop wallets that bought early in 5+ coins of the same creator). Report both versions.

- **v2.1 update (Supervisor, 6 Oct, after your round 2):** strong work. The slow-burn finding (winners keep pulling in NEW buyers for days) and "bots before $500k = bad sign" are the most useful results so far. Before the hour-4 rule goes into a paper book, it must pass three checks. Do these first, in order:

### V1.1 Price the hour-4 rule like real money
- Your exits use hourly prices and ignore liquidity pulls. On Robinhood Chain v3 coins, 78% of pools are pulled 5 to 20 h after start, and big coins usually crash about 99% before the pull. A 7-day trailing hold can be wiped out by that.
- Rerun the hour-4 rule with: real pool depth for every fill (our size $500), swap-by-swap prices (not hourly), and any pull or half-pool removal = exit at what the pool actually pays. Report mean, median, both halves.
- Add the lane test that every rule now needs: on these coins, does an outside buy of $500 get followed by a pull within 30 s more often than chance? (c10k was a honeypot that pulled on every outside buy: http://159.223.88.168/hx-c9ois/fw/fw-analyst2-status.md)

### V1.2 Overlap with the v3 rule
- How many of your 117 hour-4 trades are coins the v3 rule already bought (58 trades: http://159.223.88.168/hx-c9ois/sw/sw-a2-8-trades.csv)? What does the hour-4 rule add on coins the v3 rule did NOT buy? That added part is the real value.

### V1.3 "Bots before the gate" as a filter on the v3 rule
- You found copy bots arrive BEFORE $500k on stalled coins and AFTER on winners. On the 58 v3 trades: count bot-copy buys before the gate. Do the trades with many early bots lose more, both halves? If yes, it becomes a filter for the Coder.

- Then continue with V3 to V5 below. Price Predict is asked to give you the signed buy tape after 30 min (its signer table) and the fair coin list.

- From: the 2nd Window Supervisor, 6 Oct 2026. v2 replaces v1. Ben's brief, in his words: virality is **how a coin spreads, from the very first wallet to the end**. Map how the dominoes fall, find the point where it hits critical mass and snowballs, find what decides how far it snowballs, and find the different spread patterns. Compare winners with coins that did not make it.
- Ben's working picture (a hypothesis to test, not a fact):
  1. it starts with a certain wallet, and people follow that wallet;
  2. somebody tells a friend; you can see certain people follow other people;
  3. it hits fomo trending or DexScreener trending, or similar;
  4. a smaller account mentions it ("gem finder");
  5. towards the end, the influencers come in.
- What we already know:
  - Virality map v1 (survivor-biased, 163 coins): winners stay quiet on X early, get FRESH buyers (not recycled trench wallets), and new buyers keep coming hour after hour; influencers arrive after $1M. http://159.223.88.168/hx-c9ois/brx/virality-map.md
  - v3 runner rule (58 trades, 1.90x): http://159.223.88.168/hx-c9ois/sw/sw-a2-8-strict.md
  - Other task lists, same folder: sw-tasks-pp.md, sw-tasks-coder.md, sw-tasks-analyst1.md, sw-tasks-analyst2.md, sw-tasks-analyst3.md
- Rules: no Dune; twitterapi.io one query per question with a daily cap; Bangkok time; full addresses; paper only.

## The method (first principles)
- A coin spreads only through buyers. So the spread is a chain of buys: who bought, when, and what made them buy (they follow a wallet, a friend sent it, a board showed it, a post showed it).
- Every buy after the first can be explained by one or more "channels". Measure which channel brought each wave of buyers.
- Snowball = each wave of buyers brings in more than one new wave (like a virus with a reproduction number above 1). Find when that number crosses 1 and stays there.
- How far it goes = how many separate groups of people it reaches before the number drops below 1 again.
- Always compare: every coin that reached a low level (for example $100k), not only winners. Winners vs fizzlers, both halves of the data.

## Tasks, in order

### V1. Leader wallets and follower links (step 1 and 2 of Ben's picture)
- From the buy tape with signers (PP's day files and PP1), for every pair of wallets: how often does B buy the same coin within 60 s / 5 min / 1 h AFTER A, across many coins, compared with chance? Strong, repeated pairs = "B follows A".
- Leaders = wallets whose buys are followed by many others. Followers of followers = second hop.
- Separate the kinds: copy-trade bots (seconds), humans following a wallet (minutes), friends or groups (funded by the same source, buying in clusters).
- Output: the follow graph, the top leaders with full addresses, how many hops a typical cascade has, and whether winners have more hops / more separate groups than fizzlers.

### V2. The domino timeline per coin
- For every coin that passed $100k (13 Sep to now): one timeline of events, in order:
  - first buyers; leader wallet buys; follower waves (from V1);
  - brand new wallets vs recycled trench wallets (PP2 score);
  - trending board entries (PP7, from now on);
  - X posts by account size (PP4), first small "gem finder" post, first 10k+, first 100k+;
  - cap and buyers per 10 minutes.
- Compute the spread number per 10 minutes (new buyers this window divided by new buyers last window) and mark when it first stays above 1 (the snowball point).

### V3. Spread patterns
- Group the timelines into a few patterns by which channel came first and which carried the growth (for example: leader-wallet led, friend-group led, board led, X-caller led, crew/wash led).
- For each pattern: how often it becomes a winner, at what cap and buyer count the snowball point usually comes, and what usually stops it.
- Which early signs, visible BEFORE the snowball point, tell the patterns apart?

### V4. What decides how far it snowballs
- Among coins that reached the snowball point: what separates $5M from $20M+? Test: number of separate groups reached (distinct funding sources / new wallets), speed of new buyers, early-holder selling, narrative/stock lane, time of day, chain mood (rotation.json).

### V5. Turn it into trades (after V2 to V4)
- Every pattern or sign that looks real becomes a rule: a buy signal, a filter on the v3 rule, or a hold/sell signal. Test each on both halves, fair data only, then hand the spec to the Coder.

### Side task (small, any time): does X before the gate split the 58 v3 trades?
- Pull X posts by contract for the 24 h before each gate. Loud vs quiet, early 10k+ account vs none. Report per group, both halves.

## Hand back
- One short doc per finished task in this folder, linked from http://159.223.88.168/hx-c9ois/sw/sw-virality-status.md. V1 first. A first principles score at the end of each.

Version: v2.7 (V6 accepted; V7 hold-or-sell signal at 5x), 6 Oct 2026 15:30 Bangkok, 2nd Window Supervisor. Latest version: http://159.223.88.168/hx-c9ois/sw/sw-tasks-virality.md