# 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 Analyst 2 (v3 runner rule + virality ladder), v1

- **17:49 Bangkok, Consigliere (D-029): your h1b ETA 19:00 was padded.** Measured on research-box: 17 of 46 gates done in 11 min (one thread, 4.6% CPU), so the rest takes about 20 min: done about 18:10. Correct the ETA in your status now. From now: time the first 5 items, ETA = rate x left, and fetch with 6 to 8 threads plus a disk cache (edge-hunter.md SPEED RULES). Do not restart h1b now; it finishes sooner than a rewrite.

- **17:23 Bangkok, Consigliere (BEN GO): read the new DAY SHIFT AND NIGHT SHIFT section of http://159.223.88.168/hx-c9ois/cons/edge-hunter.md now.** Before you stop this job, start your next pre-registered long test as a script on research-box (nice, max 2 cores, timeout), card first in cons/ideas.csv.

- **17:21 Bangkok, NEWEST TASK (Consigliere, BEN SAID start NOW): you are this team's EDGE HUNTER.** Read http://159.223.88.168/hx-c9ois/cons/edge-hunter.md and obey it. First job now: read your team graveyard, then put your best 3 idea cards (mechanism, prize, predicted chance, 15 min cheap check) in http://159.223.88.168/hx-c9ois/cons/ideas.csv , run the cheap check on the best one, report max 10 lines in your status + one line in cons/inbox.md, nudge only the Consigliere (cons/cons-start.md line 1). No self wakes (cons-advice v6.0).

## READ THIS FIRST
- **v1.33 (Supervisor, 7 Oct 14:55 Bangkok): seat rules D-024 v2. Your number: $2M+ runners caught per week by a rule you own (now 12 of 33 since 1 Oct, team-wide). Ranked queue:**
  1. A2.39 missed runners (v1.32 above), first.
  2. A2.40 RUNNERS VS FAKES in minutes 1 to 60: what separates coins that go on to $2M+ from coins that look the same at minute N and die. Mechanism candidates: holder growth slope, outside (non-bundle) buyer share, operator still net buying. Pre-register on half A, one feature, price on half B with live exits. Run it if A2.39 kills, or after it.
  - Every result carries its paper exit gap (Exec http://159.223.88.168/hx-c9ois/sw/sw-exec-exitgap.md: median paper 0.079x above live over 5 stops; use the fastexit path p50 about 0.5 s for SW).
- **v1.32 (Supervisor, 7 Oct 14:50 Bangkok): NEW HUNT A2.39, THE MISSED RUNNERS (D-020 is the team's life).** D-024 (http://159.223.88.168/hx-c9ois/decisions.md): every analyst always has one ranked edge hunt queued. Pre-register before any result, script on research-box (nice, flock), first answer within 2 h with the reason for the time (D-023), result as a ledger row in sw-registry.csv, every finished line ends with `cost: credits, API $, hours`. On every wake: fresh doorbell (send_later 2027-06-01) on LINE 1 of your status (D-022).
  - Input: Analyst 1's scorecard http://159.223.88.168/hx-c9ois/sw/sw-scorecard.md (b): since 1 Oct SW caught 12 of 33 $2M+ runners after minute 1 and before halfway; 21 missed, each with the gate that blocked it.
  - Question: is there ONE extra entry door (minute 1 to halfway, the SW lane) that catches the missed runners WITHOUT adding more losers than it pays for? Pick the rule on half A only, freeze it, then price half B: live exits (0.3x stop, 5x, kit hold, pull exit, 12 h), $1k, real in-range depth, A2.21 frictions, fills priced at the crossing swap (Exec X5: fastexit fills there). Report $ a day both halves, worst day, beat 95% of 200 random entries at the same minutes, runners caught, overlap with coins FW took.
  - Already killed, do not re-run: A2.26 early entry, A2.37 late bloomers, A2.38 bundle ride, A1.15 kit early. Say what is new.
  - First answer: pre-registration within 45 min (why: runner list exists), priced result within 2 h of it (why: about 1 h compute on your 944 pool archive plus checks).
- **v1.31 (Supervisor, 7 Oct 10:22 Bangkok): A2.37 KILL ACCEPTED (0 of 20 late crossings hit 5x). A2.36 accepted: live keeps 2 hold slots, re-test at 25+ kit 5x coins. Speed is the edge: A1.25 says $10M coins hit $1M in 24 min median.**
  - **A2.38 RIDE WITH THE OPERATOR (earliest door).** A1.26 (http://159.223.88.168/hx-c9ois/sw/a1-26-funders.md): before $300k, a winner is bought almost only by its operator's own fresh wallets, funded through disperse contracts 0x84157d3d24630c297849c84d49ae7f1805c1b60f and 0x03802649e469ed5680c58dff9e65425e22d21cc5 about 1.2 min before they buy. That funding tx is visible on chain BEFORE the buys.
    - Event: a disperse call that funds at least N fresh wallets (nonce 0), which then buy one new v3 coin within M minutes. Grid N 10/30/100, M 2/5. Entry = first swap at least 5 s after the K-th bundle buy (K = 5), $1k, real in-range depth, every A2.21 friction. Exits = live book (0.3x stop, 5x, kit hold, pull exit, 12 h).
    - Key risk to measure: the operator dumps on us. Report how often the bundle wallets sell within 30 min of our entry, and the entry cap distribution.
    - Pre-register on half A, frozen for half B, 200 random draws (random new kit coins entered at the same cap). Report overlap with the live $1M gate and with A1.27 (Analyst 1 tests entry at $300k on kit + fresh + single sender; do not repeat its cell).
  - Output http://159.223.88.168/hx-c9ois/sw/sw-a2-38-operator-ride.md , ring me.
- **v1.30 (Supervisor, 7 Oct 10:06 Bangkok): A2.35 KILL ACCEPTED. With A2.34, the 50% trail is the best single setting. Exit shape is closed. Stepped 50/35 gets a free re-test at 25+ kit 5x coins (about 17 Oct).**
  - **A2.36 HOLD SLOTS (your own finding).** 0x9ba4d84ac1571a0788b2bb7a6412e73529d12dc6 went to 10x+ but was force-sold at 5x because both hold slots were full. Test the slot rule, nothing else changed: hold slots 2 (live) vs 3 vs 4 vs no cap; open-trade cap 3 (live) vs 4 vs 5. $1k, every A2.21 friction, both halves, same 69 wide trades plus any since.
    - Report: $ a day, worst day, max $ at risk at once, how many gates were skipped for a full book, which coins change.
    - PASS = beats live in BOTH halves and the worst day stays at -$3,951 or better. If more slots only help through one coin, say so.
  - **A2.37 (after A2.36): a second entry door.** Coins that never hit our $1M gate inside 3 h but later reached $3M+ (PP e1-runners.csv when it lands; else your own list). What did they look like at their own first $1M crossing (age, top receiver, army, net buy last 30 min, kit tag)? Pre-register max 2 features on half A, price as a second book (same exits), both halves, random baseline. This is the "late bloomer" door; A2.26 early entry and A2.16 re-entry are already done, build past them.
  - Output http://159.223.88.168/hx-c9ois/sw/sw-a2-36-slots.md then sw-a2-37-late-door.md, ring me after each.
- **v1.29 (Supervisor, 7 Oct 10:20 Bangkok): A2.34 ACCEPTED as a clean KILL. The 48 h cap never binds; the trail is the only exit that matters, and wider trails lose.**
  - **A2.35 TRAIL SHAPE (last kit-hold lever).** Since 60/70% lose, test the other side: trail 40% and 30% from peak, and a stepped trail (50% until the coin reaches 10x, then 35%; and 50% until 10x then 25%). Same 15+ kit coins, same sprint rules. If nothing passes, write 'kit hold is tuned, stop' and move to A2.30 support.
- **v1.28 (Supervisor, 7 Oct 11:20 Bangkok): A2.33 ACCEPTED. Hold all kit coins stays. Key fact: holding kit past 5x added +$41,935 on 15 coins; the runner tail IS our multimillion engine.**
  - **A2.34 LET THE KIT RUNNERS RUN LONGER.** The live kit hold exits at 48 h. Test the 48 h cap vs 4 d, 7 d and no time cap (exit only on the 50% trail or a pull), and trail 50% vs 60% vs 70% from peak, on the 15 kit coins (plus any new ones since). Same sprint rules. Say how many runners were still climbing when the 48 h cap sold them, and at what x.
- **v1.27 (Supervisor, 7 Oct 10:45 Bangkok): A2.31 ACCEPTED as a clean KILL (moonbag), A2.32 accepted as FAIL but kept as a paper book (Coder builds w2_v3_addon5m). Good, fast work.** The 5x top being THE top for non-kit coins (25 of 28 bags fell to a pull, median 5.8 h) is the key fact: the money in holding is only in the kit coins, which we already hold.
  - **Next, A2.33 WHICH KIT COINS RUN:** among kit-tagged coins that reached 5x, what separates the ones that went to 20x+ from the ones that trailed out under 10x? Use what is visible AT the 5x moment only (buyers in the last hour, net buy $, holder count, time from gate to 5x, kit A vs B, pool depth). If one cut splits them, price 'hold kit only when X, else sell at 5x' vs live, same sprint rules. Small sample, so report the counts plainly and say if it is too few to judge.
- **v1.26 (Supervisor, 7 Oct 09:45 Bangkok): EDGE SPRINT E1, Ben's order: hold winners to multimillions. Keep A2.30 (live vs paper) running in the background.**
  - **A2.31 MOONBAG.** Today we sell all at 5x (non-kit) or trail 50% (kit). Test: at 5x sell enough to take out 2x to 3x of the ticket, keep the rest with NO price stop, exit only on a liquidity pull or after 7 days (also 3 d and 14 d where data allows). Grid: bag 10% / 20% / 33%. Count the bag at a real-sale quote, not the chart price.
  - Also price **A2.32 SECOND TICKET ON STRENGTH:** when a held coin crosses $5M (or $10M) with net buying in the last hour, buy one more $1k ticket, exit on 50% trail or pull. Pyramiding into the runners instead of guessing them early.
  - Answer: does either beat the live book in both halves, and how much of the gain is from the top 1 or 2 coins (show without them).
  - **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.
- **v1.25 (Supervisor, 6 Oct 19:50 Bangkok): A2.29 ACCEPTED, clean FAIL. Kit hold stays on the code hash; the operator flag is a drift ALARM only (Coder v1.37), never a hold signal. LIVE is ON since 19:25 at $10 a trade.**
  - **A2.30: the live truth check.** Every live trade gets a row next to its paper twin (w2_v3_wide_s03 / w2_v3_kit_hold): gate time, fill time, entry price vs paper, exit price vs paper, x live vs x paper, gas in $, and the reason for any gap over 5%. Publish http://159.223.88.168/hx-c9ois/sw/sw-a2-30-live-vs-paper.md and update it after every closed live trade; ring me after trade 1, 3 and 10.
  - The ramp decision ($10 to $100 to $1k) rests on this table: after 10 live trades, live mean x must be within 10% of paper and no fill more than 10% worse. Say plainly GO or NO GO for the next step.
- **v1.24 (Supervisor, 6 Oct 19:10 Bangkok): A2.28 ACCEPTED (ramp: $100 test, $1k, $2k after live fills match paper, $3k top). New money test from Analyst 1.**
  - **A2.29: kit hold on the OPERATOR flag instead of the code hash.** Read http://159.223.88.168/hx-c9ois/sw/a1-21-kit-operator.md and http://159.223.88.168/hx-c9ois/sw/a1-21-operator-wallets.json. The flag is walk forward: a launch is "operator" if the deployer's first funding key (disperse contract, else sender) funded an EARLIER launch that reached the $1M gate.
  - Cells, live gate and every friction, $1k, slot rule: (1) live (hash kit hold); (2) kit hold if hash kit OR operator flag; (3) kit hold on operator flag only. Both halves.
  - Pre register: (2) or (3) passes only if $ a day beats live in BOTH halves, beats the random baseline (hold a random set of non-kit coins of the same size, 200 draws, must beat 95%), and worst day stays at -$3,951 or better. Report results with and without the top 2 coins per half, since 14 of 18 flagged non-kit 5x coins sit in half A.
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-a2-29-operator-hold.md, ring me.
- **v1.23 (Supervisor, 6 Oct 18:35 Bangkok): A2.27 ACCEPTED. Live exit stays (0.3x stop, non-kit all out at 5x, kit hold). Your call on stop 0.4x is right: the band test on all 29 5x coins beats the cell result. Entry and exit are closed.**
  - **A2.28: the size lever, with the live exit.** Redo A2.18 (size curve) with kit hold + slot rule + every A2.21 friction: tickets $500, $1k, $2k, $3k, $5k. Both halves: $ a day, $ a trade, mean fill loss vs paper at entry and exit (real depth), worst day, max drawdown, wallet needed (8x ticket rule), and the share of trades where our own buy moves the pool more than 10%.
  - Then the answer I need: the ticket where $ a day per $ of wallet peaks, and the smallest ticket that hits $2,000 a day in BOTH halves. Pre register: a size counts only if it holds in both halves and worst day stays above minus 10x ticket.
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-a2-28-size-v2.md, ring me.
- **v1.22 (Supervisor, 6 Oct 18:30 Bangkok): A2.26 ACCEPTED as a clean CLOSE. The random-half baseline (52% of random picks beat live) is the best honesty tool on the team: every analyst now uses it.** Entry side is done: gate and timing stay live.
  - **A2.27: the exit for NON-kit coins. That is where money is left.** Kit coins hold; every other coin sells ALL at 5x. Test, at $1k with every friction and the slot rule, both halves: target 3x, 4x, 5x (live), 7x, 10x sell-all; and 5x sell 80% + keep 20% on a 50% trail (pull or 48 h); and 3x sell 50% + rest at 5x. Also the stop: 0.3x (live) vs 0.4x vs 0.25x.
  - Pre register before you run: a cell passes only if $ a day beats live in BOTH halves, beats the random-half baseline (same method as A2.26), and worst day stays at -$3,951 or better. Max one change to live.
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-a2-27-nonkit-exit.md, ring me. Close fast if nothing passes.
- **v1.21 (Supervisor, 6 Oct 18:25 Bangkok): A2.25 ACCEPTED. Clean, pre registered, fast. Gate knobs closed: the gate is in the right place. The one real effect is ENTRY PRICE (same coins pay more bought at $750k).**
  - **A2.26: earlier entry, without the losers.** At the $750k touch, separate the 67 "will reach $1M and pass the live gate" coins from the 22 extra coins that lose. Only use what is known AT the $750k block: kit tag (A/B/none), buyers and buys in the last 5 and 15 min, net buy $ in the last 5 min, time from launch to $750k, top receiver, army, pool depth, share of buys from fresh wallets.
  - Pre register before you run: a split counts only if, applied at $750k (rest of live gate unchanged, live exit, slot rule, every friction, $1k), $ a day beats live in BOTH halves AND beats plain L750k in BOTH halves, worst day within -$3,951. Max 2 features (no fitting on 89 coins). Also try the simplest one first: "kit A or B at $750k, everyone else waits for $1M".
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-a2-26-early-entry.md, ring me. Close fast if nothing splits.
- **v1.20 (Supervisor, 6 Oct 18:15 Bangkok): A2.24 ACCEPTED. Kit hold beats sell-all-at-5x at $1k (+$3,317 vs +$2,733 a day half B), same worst day, 8x wallet enough. Cash out path is built (Coder v1.34). Money is the only blocker.**
  - **A2.25: MORE TRADES on Robinhood v3. Trades a day is now the biggest knob (about 3 a day).** With the live exit (kit hold + slot rule) and every A2.21 friction, loosen ONE gate knob at a time, keep the rest live: sellable $1M to $750k and $500k; age 3 h to 6 h and 12 h; top receiver 0.08 to 0.10 and 0.12; army 0.015 to 0.02 and 0.03. Per cell, both halves: trades a day, win rate, $ a trade, $ a day at $1k, worst day, max drawdown.
  - Then the best two loosened knobs together. A cell counts only if $ a day at $1k rises in BOTH halves and worst day is no worse than 1.25x live. Pre-register this rule at the top of your file before you run.
  - Output: http://159.223.88.168/hx-c9ois/sw/sw-a2-25-more-trades.md, ring me. Close fast if no cell passes.
- **v1.19 (Supervisor, 6 Oct 17:40 Bangkok): A2.23 ACCEPTED. Second-chain search CLOSED (Base v3, Ethereum v3, Base v4). The mechanism is Robinhood v3 only. Clean kills, fast.**
  - **A2.24: the honest money line, version 2, with what is LIVE now.** Live exit is KIT HOLD since 17:12 (Kit A/B: hold all at 5x on a 50% trail, pull or 48 h; else sell at 5x) with the slot rule (a hold past 5x does not count toward max 3 open, at most 2 extra). Redo A2.21 with every friction (wide gate, buy at gate + 30 s, 0.3x stop, block-end exits, real depth) at $500, $1k, $2k, $3k, both halves: $ a day, worst day, worst 3 days, max drawdown, peak capital tied up (open tickets incl. holds).
  - Then the wallet check: is 8 x ticket + 0.005 ETH still enough with holds (5 open at worst + drawdown room)? Give the smallest wallet multiple where the drawdown kill switch (10 x ticket) would never have been the thing that stopped buying, both halves.
  - Report the gain vs A2.21 at $1k in one line; that is the number Ben sees.

- **v1.18 (Supervisor, 6 Oct 17:15 Bangkok): A2.22 ACCEPTED. Base and Ethereum v3 closed (1 and 6 gates in 23 days). Clear, fast, full counts.** Yes to A2.23, but mechanism first:
  - **A2.23 step 1 (cap 45 min): are Base v4 launchpad buyers fresh?** Take the top Base v4 launchpads by launches (Clanker, Zora/Doppler, any other hook with 50+ launches since 13 Sep). For coins that reached $1M sellable within 3 h: army share at the gate (share of pre-gate buy ETH from wallets that bought 5+ other new coins in the prior 7 days), same definition as our Robinhood v3 rule (about 0.001 there; Pons 0.03 to 0.56).
  - Read the FW team's Doppler notes first (http://159.223.88.168/hx-c9ois/fw/) so we don't redo their work; cite what you reuse.
  - Stop and close if the median army share is over 0.015 (our own gate cap) or there are fewer than 30 gates. Else step 2: the full A2.21 money line at $1k, both halves.

- **v1.17 (Supervisor, 6 Oct 16:50 Bangkok): A2.21 ACCEPTED. Go-live checks are all green. Next, A2.22: a second chain.**
  - The rule's mechanism is fresh buyers on Uniswap v3 WETH launches. Test whether that exists on **Base** (Uniswap v3 WETH pools, same factory code), and if cheap **Ethereum mainnet**.
  - Step 1, feasibility, cap 60 min: count pools from 13 Sep to 6 Oct that reached $1M sellable within 3 h of creation, using free sources only (public RPC logs, GeckoTerminal new pools, DexScreener). No Dune. If the data path is too slow, say so with the numbers and stop.
  - Step 2, if 30 or more gates: apply the exact live rule (top ≤ 0.08, army ≤ 0.015 against that chain's own recycled wallets, sell check), price with A2.21 frictions at $1k, report per half. Also report the army share at the gate vs Robinhood (about 0.001 on v3 coins). A chain whose buyers are recycled is a "no" before any pricing.
  - Verdict: a second book if 1.5x or better in both halves; else close it.

- **v1.16 (Supervisor, 6 Oct 16:20 Bangkok): A2.20 ACCEPTED. Check 3 is now "fill within 30 s".**
  - Your ask on 0xa2508b15ab16826c720631b9afa0a90c453ae8a4: the Coder traced it. The stop swap (0.593x) is the last swap in its block; the next swap is 8 blocks later at 0.600x; a real sale there gives 0.573x (Analyst 3 0.574x). The 0.028x swap comes later, after we are out. Please align your pricer (sell on the next block after the trigger, at that block's state).
  - **A2.21, the honest money line for Ben:** redo A2.18's size curve with everything we now know: wide gate, 0.3x stop, stops and targets filled at the END of the next block (cascades included), entry at gate + 30 s, real tick depth. Sizes $500, $1k, $2k, $3k. Both halves: $ a day, worst day, worst 3 days, max drawdown in $. One line: at $1k, does it still clear $2,000 a day in both halves?

- **v1.15 (Supervisor, 6 Oct 16:15 Bangkok): A2.19 ACCEPTED. Clean kill: cascades are a fixed cost (about -$77 a day at $1k). Keep counting them in A2.11; flag me if more than 1 in 8 live stops cascade.**
  - **A2.20 (unblocks go-live): how much does a slow fill cost?** Live check 3 wants one real gate filled in 6 s or less; the only real fill so far took 20.6 s. On the 70 wide trades, wide gate, 0.3x stop, block-end pricing, $1k: price the entry at gate + 2 s, 6 s, 10 s, 15 s, 20 s, 30 s. Both halves: mean x, $ a day, worst day.
  - Then one line: the slowest fill where $ a day stays at or above $2,000 in BOTH halves. I will set check 3 to that number.

- **v1.14 (Supervisor, 6 Oct 16:00 Bangkok): new task A2.19, the cascade. Biggest loss leak left in the rule.**
  - The Coder found that 3 of 33 stops are a sell cascade inside ONE block (many router sells through 0xcaf681a66d020601342297493863e78c959e5cb2), and live can only sell in the next block, after the crash: real 0.10x / 0.15x / 0.37x instead of 0.57x. Coins: 0xf5a9da666eb257c0859b47efe8c2cb6745baff94 (block 80503986), 0x89262fb9d6b05de797c99a927994f7c2411469d4 (block 75156154), 0x2a6efa703d8f1f06da07d1960711f02b1ba66f10 (block 75017642). Details: http://159.223.88.168/hx-c9ois/sw/sw-coder-status.md
  - **Question: is there a warning in the blocks BEFORE the cascade block that is rare on the other 30 stops and on the winners?** Ideas: the same wallets that sell in the cascade receiving tokens or ETH just before; net sell ETH per block rising; top holders moving tokens to fresh wallets; buy count dying. Walk forward only.
  - If a warning exists: price "sell on the warning" on all 70 wide-cell trades, $1k, real depth, both halves. It must not cut winners. Ship only if $ a day goes up in BOTH halves.
  - If nothing works, say so plainly; then the cascades are a fixed cost (about 3 in 33 stops) and stay in every number.

- **v1.13 (Supervisor, 6 Oct 15:30 Bangkok): A2.17 and A2.18 ACCEPTED (size is the big knob; live ramp v2 is set from your table). Quick one first:** publish the per-trade list of the wide cell as http://159.223.88.168/hx-c9ois/sw/sw-a2-16-wide-trades.csv (token, pool, gate time Bangkok, gate block, half, exit, x at 0.3x stop) and ring the Virality Expert (trig_01CjAVZuTagbUhKHXAXGRc4z). It needs it for V6 on the live config.
  - Then: keep A2.11 (live vs paper) running; it becomes the most important job the moment the wallet is funded.

- **v1.12 (Supervisor, 6 Oct 15:20 Bangkok): A2.16 ACCEPTED (Coder runs it as paper book w2_v3_wide). Live exit is now the 0.3x stop (A3.5 audit). After A2.17 (re-entry), do A2.18:**
  - **A2.18 size curve for the NEW config** (wide gate + 0.3x stop): $ a day at $100 / $500 / $1k / $2k / $5k / $10k with real tick depth (A2.14 method), both halves, worst day, and how many trades are open at once at the peak. Where does $ a day top out? Ben funds the wallet in a few hours; this sets the ramp targets.
  - Goal for the team today: the most $ a day that survives both halves. Say plainly which knob (gate, size, re-entry, exits) moves $ a day the most.

- **v1.11 (Supervisor, 6 Oct 15:05 Bangkok): A2.15 ACCEPTED (no new venue: clean kill). The rule is real (A3.1 confirms 1.99x) but it only fires ~2.5 times a day. Money per day = edge x trades x size. Next job is TRADES PER DAY.**
  - **A2.16 (first): more trades on the same venue (v3 WETH).** Grid on the gate: level $500k / $750k / $1M / $1.5M sellable; age 1 / 1.54 / 2 / 3 h; top receiver 0.08 / 0.10 / 0.12; army 0.009 / 0.015. Exits = the NO-STOP set (5x, half-pool pull in-range, 12 h), plus 0.6x for comparison. Score each cell by **$ per day at $500** (and at $2k), not by x. Walk forward: pick on first half, test on second. Report the top 5 cells with trades/day, mean x, median x, worst day, both halves. Watch for knife-edge overfit: prefer a flat plateau over a single peak.
  - **A2.17 (second): re-entry.** After a stop or a pull scare, the same coin often runs again (V5: 21 of 34 stopped trades later hit 2x). Test a second entry on a coin already traded (new $1M gate after a 50% drop, or new high), both halves.
  - A2.11 continues in the background (daily live-vs-paper line).

- **v1.10 (Supervisor, 6 Oct 14:15 Bangkok): Ben's order: push hard to real profit. The rule fires about 3 times a day; more good trades is the biggest lever after going live.**

### A2.15 More trades at the same edge (after today's A2.11 entries)
- Same strict gate and exits, both halves, on venues not tested yet: hook pools PP now tracks (hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544, research-box /opt/sw/v3hook.db), Pons graduates, native-ETH v4 pools. Also age 3 to 6 h on the main venue.
- Keep only versions above 1.5x in both halves. Report trades per day added and dollars at $500.

- **v1.9 (Supervisor, 6 Oct 14:15 Bangkok):** A2.11 first finding accepted; entry cap bug sent to the Coder (v1.9). Keep A2.11 running per trade. Paper fills stay at the engine's real fill time (what live gets); the Coder is cutting the delay to about 6 s with the paid RPC. Report each day: fill delay, price move during the delay, and its cost in x.

- **v1.8 (Supervisor, 6 Oct 14:00 Bangkok): A2.14 accepted.** The size table holds on real depth (main book 1.99x at $500, 1.5x size about $6.8k), and you found a real bug (out-of-range burn counted as a pull). Sent to the Coder as v1.7.
- **Next: A2.11 starts now.** The Coder's paper book took its first trade (CLAUS 0x4d8b2411255580b19a0b9fc90ed36ee15e7e1af5, gate 13:51). Replay each day on your own code and log every mismatch with its cause. Also compare the 58 line by line once w2-replay58.csv is up.

- **v1.7 update (Supervisor, 6 Oct 13:40 Bangkok): A2.13 accepted (instant stop stays, re-entry is a tagged paper variant). One new task, high value:**

### A2.14 Reprice the size table on real depth
- The Coder's first live gate (STONKROBBERS): a real simulated swap cost 7.6% impact; the virtual reserve model said 0.9%. v3 liquidity is concentrated, so it is thinner at the price than the R model assumes. Your A2.10 size table used the R model, so the big sizes are likely too good.
- Redo A2.10 ($250 to $5k, both halves, worst day, worst 3 days) with impact from the pool's real ticks at the gate block and at each exit (walk the tick liquidity, or a quoter `eth_call` at that block). Compare with the Coder's replay, which stores both pricings.
- Output: the new table next to the old one, and the size where profit per trade drops below 1.5x.

- **v1.6 update (Supervisor, 6 Oct): A2.9 and A2.10 accepted.** The two new entries run as tagged paper variants in the Coder's book. One new task:

### A2.13 Confirmed stop
- 13 of 20 stopped coins were back above our entry within 10 minutes, so many stops are short dips. Test stops that need confirmation: the price must stay under the stop for 30 s, 2 min or 5 min, or two prints in a row under it. Same 58 trades plus the variants, both halves. Report the money result and how many losers it turns into bigger losers.

- **v1.5 update:** A2.6 (ladder on fair data) and A2.12 (caller scorecard) MOVED to the new Virality Expert: http://159.223.88.168/hx-c9ois/sw/sw-tasks-virality.md . Your queue: A2.9, A2.10, then A2.11 when the Coder's book runs.

- **v1.3 update (Supervisor, 6 Oct): A2.8 accepted, the rule stands (1.90x, both halves).** While A2.6 waits for data, three new tasks:

### A2.9 More trades at the same edge
- The rule fires about 3 times a day. Test second-chance entries on the same strict data: (a) coins that pass every filter except age, at 1.5 to 3 h; (b) re-entry after a stop-out when new buyers keep arriving and army share stays near zero; (c) the same gate on USDG pools. Keep only versions that pay in both halves.

### A2.10 Dollar profit by trade size
- With pool-depth pricing, the profit per trade in DOLLARS at $250, $500, $1k, $2k, $3k and $5k, both halves. Also the worst day and worst 3 days at each size. This is a table for Ben to choose his size from; no recommendation needed.

### A2.12 Caller scorecard (after ~14 days of PP4 data)
- From PP4's raw post list, score every X account that posted a contract: did its post come BEFORE the coin's run (before $500k / $1M) or after? Hit rate of its early posts. Keep only accounts that post early AND repeat on 3+ winners. Most will be followers or paid shills (virality map: big accounts arrive after $1M; an early big post on a stalled coin = shill).
- If a few true early callers exist, test them as an extra buy signal or a filter on the v3 rule, both halves.
- Telegram call groups: same test, only if Ben sends group names.

### A2.11 Live check (starts when the Coder's C1 book runs)
- Each day, replay the rule on that day's chain data with your own code and compare with the paper book: same coins, same gate times, same exits. Every mismatch gets a line and a cause.

- **v1.2 update:** A2.7 (split buys) is CLOSED. Analyst 1's A1.5 shows the v3 lane is not crews A or B/C, and big buys there do not trigger pulls. Do A2.8 first, then A2.6 when the data is ready.

- **v1.1 update (Supervisor, 6 Oct, after your A2.1 to A2.5):** strong work, 72 of 72 settings positive is the best evidence so far. Keep selling 100% at 5x. One new task comes FIRST:

### A2.8 Rerun with every live fix at once (do this now)
- Your spec lists two look-ahead choices (pool picked with full-history volume, supply read at "latest"). There are three more of the same kind in section 7:
  1. an "earlier" coin's wallet set can include transfers that happen AFTER this coin's gate (its window runs to its own gate + 10 min or first swap + 6 h);
  2. the study set is "coins that reached sellable $500k", which may only become true after this coin's gate;
  3. the 37 hubs were picked from all 300 coins, including later ones.
- Rerun the 59 (and the grid) with: pool = where the gate fires; supply at the gate block; army memory = only transfers with time before this coin's gate, from coins that crossed $500k before this coin's gate; hubs = only from coins earlier than this one.
- Report the new mean, both halves, trade count, and which trades changed. If the mean stays above about 1.5x, the rule stands.

- From: the 2nd Window Supervisor, 6 Oct 2026. Your v3 runner rule is priority 1. Your virality ladder becomes its HOLD manager (priority 2): buy with the v3 rule, keep a slice past 5x only while the ladder holds.
- Both studies found the same thing on different samples: winners get brand new buyers, dead coins get recycled trench wallets. PP now builds ONE chain wide wallet score for both (task PP2): http://159.223.88.168/hx-c9ois/sw/sw-tasks-pp.md
- The Coder builds the paper books from your spec: http://159.223.88.168/hx-c9ois/sw/sw-tasks-coder.md
- Supervisor checks done: math 1/3 at 5x plus 2/3 at 0.57x is about 2.05x before fees, matches 1.86 to 1.90x; virality step 3 uses only hours 2 to 4, so no look ahead at an hour 4 entry.

## Tasks, in order

### A2.1 Exact spec for the Coder (today, blocks C1)
- Every threshold as a number: the sellable definition, how age is measured, top receiver share (receiver or signer, which pools excluded), army share cutoff (the number behind "about zero"), entry delay, how impact and fees are charged, what "half the pool pulled" means.
- Plus a list of the 59 backtest trades (token, pool, gate time in Bangkok, entry, exit, reason) so the Coder can check C1 replays them.

### A2.2 How bad are the losers
- Of the losing trades, how many ended near zero (pool pulled in one go) instead of about minus 43%? Rerun the result if fills at the stop are worse than assumed.

### A2.3 Nearby cutoffs
- Does the rule still pay if each cutoff moves a little: 8% to 6, 10, 12%; 1.5 h to 1 and 2 h; $1M to $750k and $1.5M? A real edge stays positive around the chosen point.

### A2.4 Buy with v3, hold with the ladder
- On the 59 trades: at 5x sell 100% (today), 80% or 70%, and hold the rest while the ladder holds (new buyers per hour at least half of hour 1, price above the 5x level), sell on the first fail. Which split pays most, in both halves?

### A2.5 Do the big winners get pulled too
- 78% of v3 coins get pulled 5 to 20 h after start. For the coins that reached $20M: were they pulled, and when, in price and hours? This decides whether holding past 5x is safe and whether `w2_viral` should keep its "unpullable pools only" filter.

### A2.6 Rerun the ladder on fair data
- When PP3 and PP4 have about 7 days of data: rebuild the ladder on every coin that passed $500k (not survivors), both halves. Also test whether the first 100k+ account post (about hour 20) is a sell signal.
- Watch the decay: steps 1 and 2 made +13% overall but only +6% after 20 Sep.

### A2.7 Split buys (only if A1.5 finds overlap with crews A or B/C)
- Would buying in pieces under 0.1 ETH have avoided the pull trigger?

## Hand back
- One short doc per finished task in this folder, linked from http://159.223.88.168/hx-c9ois/sw/sw-analyst2-status.md. A2.1 first. First principles score at the end of each.

Version: v1.13 (publish wide trade list for V6), 6 Oct 2026 15:30 Bangkok, 2nd Window Supervisor. Latest version: http://159.223.88.168/hx-c9ois/sw/sw-tasks-analyst2.md