# Analyst 1 status archive (older entries, newest first)

- **10:09 Bangkok 7 Oct: A1.25 done. A $10M coin is a kit coin that gets to $1M fast and quiet, bought first by brand new wallets. The pre-registered rule is a clean KILL.** http://159.223.88.168/hx-c9ois/sw/a1-25-anatomy.md , table: http://159.223.88.168/hx-c9ois/sw/a1-25-anatomy.csv
  - v3 WETH since 13 Sep: 27 coins reached a fair $10M (17 in A, 10 in B), 3 reached $30M+, 89 reached $1M and died under $3M. In half B all 10 winners are kit coins (0 of 11 non-kit).
  - Top separator on half A: 77 sells or fewer before $1M (lift 1.83 A, 2.04 B). As a hold rule: A $5,529 vs live $5,583 a day, beats random 68%; B flags 0 non-kit coins. KILL. It mostly re-finds kits, which the kit hold already holds.
  - Creators never repeat (0 of 27 winners). Window 1 crews: 0 matches.
  - Pre-registered for half C (20+ kit trades): fresh-buyer kit hold (kit coins whose pre-$300k buyers have median nonce 1 or less keep the kit hold; other kit coins exit at 5x). Inside kits: 8 of 22 vs 1 of 11 (A), 8 of 22 vs 2 of 13 (B).
- **09:50 Bangkok 7 Oct: A1.25 started (multimillion anatomy). PRE-REGISTRATION, fixed before I look at any result:**
  - Coins: Uniswap v3 WETH pools 13 Sep to now (v4 caps are polluted by wash and no-log clones). Winner = peak cap 10M dollars or more. Dead = reached 1M, peak under 3M.
  - Features: only what is known at the FIRST 300k moment and at the FIRST 1M moment (buyers, funders, crew badge, creator and funder history, kit hash, time to get there, ETH in pool).
  - Pick: separators are ranked by lift on half A only (to 30 Sep). The top one with 5+ winners in half A becomes the rule. Entry rule if it is visible at 300k, hold rule if at 1M.
  - Test once: $1k at real depth, A2.21 frictions, both halves, vs the live book (w2_v3_wide_s03 + kit hold) and vs 1,000 random picks of the same count. PASS = both halves beat the live book AND beat 95% of random. Anything else = KILL, written up today.
- **Daily edge row (A1.23 job, 00:15 Bangkok, one line a day):**
  - 2026-10-06: paper 1 trades, all-out $3776.19, kit hold $8769.43, gates 5, new buyers/trade 1839 | live 0 trades $0 | 7-day: $3765 a day, gates 7.4 a day, buyers 1074 | no alarm
- **09:40 Bangkok 7 Oct: A1.24 done. Engine matches replay: 0 of 7 gates missed since it started (6 Oct 13:04).** http://159.223.88.168/hx-c9ois/sw/a1-24-trade-count.md
  - Low count = engine started 12:58 on 6 Oct (missed HOODCHAN, STONKROBBERS, SS before it existed), CLAUS passed before the wide book existed, and the CHONK live buy died on the lowercase address bug (already fixed by the Coder). 7 Oct is quiet: 2 gates in 9 h, 1 pass.
  - Gates per day flat (about 9), receiver fails about 1 in 3, same as before. No filter change.
- **00:40 Bangkok 7 Oct: first nightly run worked (00:15, no alarm). Two fixes to my own job:** 6 Oct was frozen as a part day (2 trades, 2 gates at 20:40), so it now comes from the full-day paper books (1 trade, 5 gates); and the status line for a day is now replaced, not repeated. 7-day: $3,765 a day, 7.4 gates a day, 1,074 buyers per trade. Job md5 186bfd4c9e811789147ddd293885bb90.
- **20:50 Bangkok: A1.23 done. Daily edge row + alarm is live (00:15 Bangkok cron, first run on today's data done, no alarm).** http://159.223.88.168/hx-c9ois/sw/a1-23-daily-alarm.md
  - Fixed thresholds: 7-day all-out $ a day < $2,000 for 3 days; gates 7-day mean < 4; new buyers per trade 7-day median < 500. Alarm goes to w2-live-events.md via the Coder's write_event().
- **20:40 Bangkok: A1.22 done. The edge is FLAT, not fading.** http://159.223.88.168/hx-c9ois/sw/a1-22-edge-decay.md
  - $ a day (all out, $1k exact): slope -$58 +- $244 a day over 16 trading days (p 0.81), +$696 +- $840 over the last 7. Kit hold estimate: flat to up. No crossing of $2,000 on any fitted line.
  - New buyers per trade 2 h after the gate are RISING (+106 a day, p 0.02). Kit launches rising (Kit B 9 to 50 a day).
  - Watch: $1M gates per day in the last 7 days slope -1.0 a day (p 0.11, today partial).
- **19:01 Bangkok: A1.21 done. A hash-blind operator rule finds Kit B, the 3,587 family and Kit A from their first launches.** http://159.223.88.168/hx-c9ois/sw/a1-21-kit-operator.md , wallets: http://159.223.88.168/hx-c9ois/sw/a1-21-operator-wallets.json
  - Rule: key = the deployer's first funding (contract if via one, else sender). Flag if that key funded an earlier coin that reached the $1M gate. Walk forward: Kit B 136 of 140, 3,587 family 16 of 17, Kit A 79 of 107. Flagged launches reach the gate 24.7% vs 0.47% for the rest.
  - Operator B = disperse `0x84157d3d24630c297849c84d49ae7f1805c1b60f`: one operator rewrote its kit 4 times (25 Sep to now, 155 launches, 34 gates).
  - Operator A = 24 funders that also launched 129 coins through the plain position manager (20 gates, 16 of our wide trades).
  - Needs a money test: flagged non-kit coins run like kits after 5x (median 4.76x vs 1.59x, 18 coins, mostly half A).
- **18:46 Bangkok: A1.20 done. No Kit C passes.** http://159.223.88.168/hx-c9ois/sw/a1-20-kit-c.md
  - No launcher with 20+ launches besides Kit A and Kit B has 10+ coins reaching 5x in both halves.
  - Watch: the 3,587-byte kit family (17 launches 29 Sep to 3 Oct, 10 reached the $1M gate = 59%, 6 wide trades mean 2.61x). Too few to test and quiet since 3 Oct. Tag it, and test it if it comes back.
  - Kit edge is the tail: top 10% peak after 5x, Kit B 10.4x vs non-kit 3.3x (half B). The median gap is small (Kit A 1.44x of non-kit).
  - Skip candidates (77% to 100% pulled within 2 h, never reach our gate): launcher `0x9b226f198ef0038d054be6265aa9f76c0d6e5e78` and 4 token templates. Tracking only.
- **17:24 Bangkok: A1.19 done. Who buys pulled coins: nothing splits, close it. No gate change.** http://159.223.88.168/hx-c9ois/sw/a1-19-pull-source.md
  - The 490 instant pulls (rebuilt, a median 19 s after the first swap) are bought by old sniper wallets (84% / 55% of buy ETH) through small sniper contracts, with 0% disperse funding. Our gated coins have none of that, so the gate already keeps them out.
  - 49 of the 63 launches by `0x9b226f198ef0038d054be6265aa9f76c0d6e5e78` were instant pulls, and 0 reached the gate.
  - The 5 gate pulls look like our winners (fresh disperse-funded wallets). Skip filters on direct transfer or disperse `0x03802649e469ed5680c58dff9e65425e22d21cc5` all lose in half A at $1k (best +$2,342 vs +$2,733 a day).
- **17:10 Bangkok: A1.18 done. The fresh buyers are brand-new wallets funded through disperse contracts minutes before they buy. No gate filter from it pays.** http://159.223.88.168/hx-c9ois/sw/a1-18-buyer-source.md
  - Routes: a fee router (`0xf2a277490c6e1d511e236c64895a0fef39aa8098`) carried most buys until 1 Oct, then 0%. Buyers moved to Uniswap SwapRouter02. The fomo router carried 0 buys.
  - Funding: 84% to 87% of pre-gate buy ETH comes from wallets funded less than an hour before buying, mostly via disperse `0x84157d3d24630c297849c84d49ae7f1805c1b60f` or `0x03802649e469ed5680c58dff9e65425e22d21cc5`.
  - Younger buyer wallets go with winners in both halves, but as a filter it loses money (A: +$3,246 vs +$5,412 a day). Route mix tracks the date, not the coin.
- **Daily check, live paper book `w2_v3_wide_s03` vs my rebuild (one line a day):**
  - 6 Oct (15:24 to 16:25 Bangkok, since the book went live): book 0 trades; my rebuild 0 sellable $1M gates. One new pool showed a $2.65M price with almost no ETH behind it (`0x291287c3ac1ecdaff1e56e03511142cef3d8d58e`), and both correctly skipped it. MATCH.
- **16:25 Bangkok: fast-gate tilt pre-registered** at the top of http://159.223.88.168/hx-c9ois/sw/a1-17-size-tilt.md (60 min cut, half C from 7 Oct, run once at 30+ trades).
- **16:20 Bangkok: A1.17 done. No size tilt ships.** http://159.223.88.168/hx-c9ois/sw/a1-17-size-tilt.md
  - No gate feature raises $ a day in both halves without a worse worst day.
  - Closest: coins that reach $1M within 60 minutes of launch get the bigger ticket. Same total stake, $ a day up in both halves (A +$6,656 vs +$5,412, B +$4,634 vs +$4,142), but half A's worst day drops from +$223 to −$1,687. Paper tag only.
  - Farm-funded and Kit B flip between halves, so they can't be tilts.
- **16:15 Bangkok: A1.16 done. Cascade sellers have one funder, but it is not a skip signal; close cascades.** http://159.223.88.168/hx-c9ois/sw/a1-16-cascade-funders.md
  - All 60 traced sellers (20 per cascade) were funded by the Kit B disperse contract `0x84157d3d24630c297849c84d49ae7f1805c1b60f`, one caller per coin.
  - The same funder is behind the buyers of 23 of our 70 wide coins (14 of 15 Kit B plus 11 others). Those coins did BETTER in half B (2.40x vs 1.81x).
  - Skipping them would cost about $1,974 a day in half B; cascades cost about $77. Tag `farm_funded` for tracking only.
- **16:05 Bangkok: A1.15 done. Don't buy Kit coins early; keep the $1M gate.** http://159.223.88.168/hx-c9ois/sw/a1-15-kit-early.md
  - **Kit B devs cash out before $1M.** In 93% of Kit B launches the insiders (deployer plus bundle wallets) sold more ETH than they put in before $1M; the median insider took out 5.9x. Early buyers are who they sell to.
  - **No early cell beats the wide rule in both halves.** Buying at $250k with a 5x exit: Kit A +$2,698 a day, but Kit B −$698 a day (wide rule +$2,035 / +$2,982 on the same days).
  - The only big Kit B number (+$8,394 a day, buy at the first outside swap, sell at $1M) is two coins at 77x and 71x. In Kit A the same cell loses money. It's a lottery.
  - **Overlap: 30 of our 36 kit trades (83%) are also first window mc5 entries.** Cap the combined size per coin across both books.
- **15:55 Bangkok: A1.14 done (where v3 coins come from).** http://159.223.88.168/hx-c9ois/sw/a1-14-v3-source.md
  - **Gate coins come from two copy-paste launch kits.** Each dev deploys their own copy, then launches seconds later.
    - Kit A: 41% of its 107 launches reached the gate. It stopped on 1 Oct.
    - Kit B: 20% of 140 launches reached the gate. It is live, 14 to 50 launches a day.
    - Everything else: 0.6%.
  - **The flow is flat in count (about 4 wide coins a day) but now hangs on Kit B:** 15 of the last 7 days' 26 wide coins (58%). Kit B coins average 2.44x on the wide rule.
  - Kit B is funded through one disperse contract, `0x84157d3d24630c297849c84d49ae7f1805c1b60f`. The same transaction gives 0.01 ETH of gas money to fresh wallets, and those wallets buy 26% to 38% of the pre-gate ETH (the dev's bundle).
  - No second venue: the kit wallets never launch on v4 and are not active on Base, Arbitrum or BSC.
  - **rep1_pre loses money on the wide cell** (+$2,178 vs +$2,480 a day), so drop it.
  - **The 1.4 to 1.5 ETH wallets are a proven farm** of `0x3393cab3fcfc5668b46d83b89f924c01f00aa766`, which also funded Kit A deployers. Its coins still paid 1.46x on our rule, so tag them, don't skip them.
- **15:30 Bangkok: A1.13 money map done (task v1.7, scored in $ a day).** http://159.223.88.168/hx-c9ois/sw/a1-13-money-map.md
  - **Only v3 makes real money in both halves after $500 real fills.**
    - The v3 rule with its own exits: **+$1,670 a day** (A +$1,759, B +$1,601, 81% of days).
    - Plain outsiders buying in the 30 min after the $1M gate: +$1,291 a day (A +$2,013, B +$730).
  - **No v4 cell is a big edge.** Pools are shallow (Pons about 4 ETH deep, so $500 costs about 5% each way). Real outsiders lose at the median in every v4 cell. The fast cells (first 2 min in c10k, other, c500) lose $39k to $52k a day.
  - One v4 cell to watch: no-hook fee 3000 / spacing 60 after $500k. It makes +$1,776 a day, but half B is only +$150, the median coin is 0.63x, and 5 coins give 91% of the gain. Not tradeable yet; ask PP to test it on pp3-cross.
  - **getCode fix: 6 labels changed, all in A1.11.** The 6 "farm wallets" were the coins themselves (token contracts). My printing mistake. The `rep1_pre` filter result is unchanged after removing every contract: 49 coins at 2.08x. A1.9 and A1.12 labels stand (20 A1.9 wallets are EIP-7702 delegated wallets, still wallets).
- **14:23 Bangkok: A1.12 done. All three new groups are traps.** http://159.223.88.168/hx-c9ois/sw/a1-12-new-crews.md
  - `0x04f2cb48fa0c60da3a6241077f1b7c549b9b9094`: looks like a honeypot. Outsiders bought 10.74 ETH and sold 0.003 ETH. All 37 pools were pulled (median 105 s). Early $9+ outside buyers lost 100%.
  - Fee 2500 / spacing 50 class (180 pools today): 100% of the buy ETH comes from 35 repeat wallets, a wash farm like c10k. Outsiders lost 98%.
  - `0xa9707935dd88323fba9aa9d6a95e3ab82c1ba786`: 97% of the buy ETH is its own. Outsiders are at −38%.
  - PP's hook `0x6404ddb4f33b4d974e570f66288e2a7ca469a840` is a different class (fee 0, quoted in token `0x5fc5360d0400a0fd4f2af552add042d716f1d168`). Not scored yet.
  - Asks: block all three; add a live honeypot guard (outside sells vs buys).
- **14:05 Bangkok: note on the pull-rate level gap with Price Predict** (their pull_daily: B/C 19% to 23% pulled within 60 s at $10 to $50, A 5% to 11%; mine 51% to 57%). The gap is the age window, not a disagreement.
  - Mine counts only outside buys in the coin's first 120 s, which is our entry window.
  - Over all ages, my numbers drop to 2% to 3% pulled within 60 s, because late buys meet crews that have already finished pulling.
  - Price Predict's mix of ages lands in between. Both agree on the direction: B/C and A pull more after bigger outside buys. For our 10 s entry, the first-120 s number is the one that matters.
  - Task list unchanged since 13:34. A1.10 waits for the live test (pond launches paused). New crews wait for PP's day-file rebuild.
- **14:00 Bangkok: A1.11 done.** http://159.223.88.168/hx-c9ois/sw/a1-11-self-buy-v3.md
  - The v3 rule is mostly clean. In 51 of 57 coins, repeat wallets bought under 10% of the pre-gate ETH (median 0.1%).
  - 2 wallet farms pumped 6 coins (21 to 24 Sep). Each farm used fresh wallets that bought in exactly 3 coins, so the army filter (3+ earlier coins) never trips.
  - **New walk-forward entry filter:** skip if more than 5% of pre-gate buy ETH comes from wallets seen pre-gate in 1+ earlier gated coin. Result: 49 coins at **2.08x** (A 2.18x, B 2.01x) vs 1.94x for all 57. The 8 dropped coins averaged 1.08x. The $20M winners in the set score 0.0% to 0.5%, so the filter leaves them alone.
  - **Hold side:** "new buyers" in hours 0 to 4 can be repeat wallets. Coins with more than 20% repeats averaged 1.17x vs 2.14x for the rest. That's after entry, so it belongs in the virality hold check, not the entry rule.
  - Next: A1.10 when the live test buys; the new crews once PP's rebuild resumes.

## Earlier update (A1.9)
- **v1.3: A1.9 done, A1.10 waits for Ben's live test (13:50 Bangkok).**
  - **Correction:** the "non-pulling crews" edge (`0x7028bab73f82dff14b199593abb76c8b6acb3aa9`, `0x2dacdf0f44d07234832799a10a7cc5c8eff0223a`) is not real.
    - Their early "outside" buyers were the same 20 wallets in every coin of both crews (the crew's own wash). That's why no pulls followed.
    - "+464%" was my paper exit. The real result was −24% and −46%.
    - Independent buyers put in only 0.19 ETH and lost almost all of it.
    - Both crews stopped launching (last about 3 Oct and 27 Sep).
  - **Walk-forward crew ranking:** no crew qualifies. The "safe crew" flag picks 28 coins at +60% for independent buyers, vs +68% for all the rest, so no edge. A, B/C and `0x0ec66d31aab7e387b2f4cda075e39f7a6a0f541b` confirm independent buyers lose there.
  - **Window 1 has no proven real-money edge today.** The $4 ticket on A and B/C (A1.10) is the only open live question.
  - **Blocker:** no swaps after 4 Oct. The day-file rebuild on main has been paused since at least 12:47 Bangkok (disk under 12 GB), so the new crews (`0x04f2cb48fa0c60da3a6241077f1b7c549b9b9094` 37 launches, `0xa9707935dd88323fba9aa9d6a95e3ab82c1ba786` 12) can't be scored.

## Earlier update (A1.6 to A1.8)
- **v1.1 tasks A1.6 to A1.8 done (13:20 Bangkok).** New headline, and it changes the plan:
  - **A1.6:** in crews A and B/C, our buy may BE the pull trigger. Early outside buys of $9 or more were pulled within 60 s 48% of the time vs 14% expected. Real buyers of $9+ lost money. Only dust (under about $4) is let in. So the ticket should shrink, not grow.
  - ~~Crew `0x7028bab73f82dff14b199593abb76c8b6acb3aa9` lets outside money in~~ WITHDRAWN by A1.9: those buyers were the crew's own wallets, and +464% was paper.
  - **A1.7:** depth costs only 5 points at $20 and 24 at $150, so depth isn't the gap to Price Predict. If our buy forces a pull, A + B/C with the 10x plan drops to −95%. Selling all at 2x still makes +35% even then.
  - **A1.8:** the pull-time sell rule holds walk-forward at 6 cut points (+211% to +236% at 10 s, every day positive). But almost every hot crew pulls after 200 s, so the rule mostly means "all at 10x", which is unsafe on A and B/C.
- **New final rule for window 1:**
  - A and B/C: 10 s only (no 35 s fallback), ticket 0.0015 ETH (about $4), sell all at 2x.
  - ~~`0x7028bab73f82dff14b199593abb76c8b6acb3aa9` and `0x2dacdf0f44d07234832799a10a7cc5c8eff0223a`: $80 to $150~~ WITHDRAWN by A1.9.
  - Other hot crews: $20, pull-time rule.

## Earlier headline (A1.1 to A1.5)
- **All 5 tasks done.** A1.4 was scored with a walk-forward funder tag from crew.db; no need to wait for PP5's file.
- Headline:
  - The hot crew filter (5 of the crew's last 10 ended coins hit 3x) is positive on every test day: +169% a coin at 10 s, +99% at 35 s. It beats the fixed A + B/C list (+105% / +47%).
  - Each crew needs its own sell plan, based on how fast it pulls.
  - The v3 runner lane is not crews A or B/C, and big buys there don't trigger pulls (checked on Analyst 2's exact 59).
  - A launch-time tag (the deployer was funded by an earlier creator of the crew) catches 77% of A or B/C coins with 0 false tags, and trading on it at 10 s made +145% a coin.
- Go-live gate status:
  - fair crew pick positive every test day: **PASS** (5 days);
  - 35 s entry still pays: **PASS**;
  - launch tag matches the label 80%+: precision 100%, recall 77% (just under); untagged coins fall back to 35 s;
  - 30+ paper coins under the final rule: **OPEN** (needs the Coder's book `w1_hot`).

## Docs
| Task | Result | Link |
|---|---|---|
| A1.1 Fair crew pick | hot +169%/coin, all 5 days positive | http://159.223.88.168/hx-c9ois/sw/a1-1-fair-crew-pick.md |
| A1.2 Buy at 35 s | still pays: +47% (A or B/C), +99% (hot), all days positive | http://159.223.88.168/hx-c9ois/sw/a1-2-buy-at-35s.md |
| A1.3 Sell plan per crew | crew A all at 10x (+189% test); fast pullers sell at 2x or skip; 0x0ec66d31aab7e387b2f4cda075e39f7a6a0f541b = review crew 0xae3d565892c79fa8fc9243b380283dd419e28ec9 | http://159.223.88.168/hx-c9ois/sw/a1-3-sell-per-crew.md |
| A1.4 Launch tag score | recall 77%, 0 false tags; +145% at 10 s vs +47% at 35 s | http://159.223.88.168/hx-c9ois/sw/a1-4-launch-tag.md |
| A1.5 v3 overlap | not A or B/C; a 0.1+ ETH buy precedes a pull within 60 s in 0.10% of cases vs 0.37% for a random window | http://159.223.88.168/hx-c9ois/sw/a1-5-crew-overlap-v3.md |
| A1.6 Buy size vs pulls | A + B/C: no bigger safe size, $9+ buys pulled 3.5x normal; cap at about $4. `0x7028bab73f82dff14b199593abb76c8b6acb3aa9`: no pulls, size up | http://159.223.88.168/hx-c9ois/sw/a1-6-buy-size.md |
| A1.7 Depth-priced fills | depth costs 5 points at $20; trigger risk is the real danger; A + B/C sell all at 2x (+35% even if our buy forces a pull) | http://159.223.88.168/hx-c9ois/sw/a1-7-depth-fills.md |
| A1.8 Walk-forward sell rule | holds at 6 cut points, every day positive; mostly means all at 10x; not on A + B/C | http://159.223.88.168/hx-c9ois/sw/a1-8-walk-forward-sell.md |
| A1.9 Non-pulling crews | withdrawn: the "outside" buyers were 20 crew wallets in every coin; real result −24%; both crews stopped; no crew qualifies walk-forward | http://159.223.88.168/hx-c9ois/sw/a1-9-non-pulling-crews.md |
| A1.11 Self-buy check on v3 | mostly clean; 2 farms (6 coins) slip past army; new pre-gate repeat filter 2.08x vs 1.94x, both halves up | http://159.223.88.168/hx-c9ois/sw/a1-11-self-buy-v3.md |
| A1.12 New crews | all three traps: 0x04f2 honeypot, fee 2500 class wash farm, 0xa970 self-buying | http://159.223.88.168/hx-c9ois/sw/a1-12-new-crews.md |
| A1.13 Money map ($ a day) | only v3 pays in both halves: rule +$1,670 a day, outsiders after gate +$1,291; no v4 edge after $500 fills; A1.11 getCode fix (6 coin addresses mislabeled, result unchanged) | http://159.223.88.168/hx-c9ois/sw/a1-13-money-map.md |
| A1.14 v3 coin source | two launch kits make the gate coins (A 41%, B 20% reach $1M vs 0.6% for the rest); Kit B is 58% of the last 7 days of flow; rep1_pre costs $302 a day on the wide cell (drop it); the 1.4 to 1.5 ETH farm is 0x3393cab3fcfc5668b46d83b89f924c01f00aa766, tag only | http://159.223.88.168/hx-c9ois/sw/a1-14-v3-source.md |
| A1.15 Kit early buy | no: Kit B insiders sell 5.9x their buy before $1M; early cells lose in Kit B (−$698 a day at $250k vs wide +$2,982); 83% overlap with mc5 | http://159.223.88.168/hx-c9ois/sw/a1-15-kit-early.md |
| A1.16 Cascade funders | one funder (0x84157d3d24630c297849c84d49ae7f1805c1b60f) for all 60 traced sellers, but it funds 23 of 70 wide coins that did better (2.40x vs 1.81x in half B); no skip, tag only | http://159.223.88.168/hx-c9ois/sw/a1-16-cascade-funders.md |
| A1.17 Size tilt | none ships; closest is fast gate (within 60 min) up in both halves but worse worst day; farm and kit flip between halves | http://159.223.88.168/hx-c9ois/sw/a1-17-size-tilt.md |
| A1.18 Buyer source | fresh buyers = minutes-old wallets funded via disperse contracts; fee router 0xf2a277490c6e1d511e236c64895a0fef39aa8098 went to 0% on 2 Oct; no gate filter pays | http://159.223.88.168/hx-c9ois/sw/a1-18-buyer-source.md |
| A1.19 Pull buyers | instant pulls = old snipers via sniper contracts, never pass the gate; gate pulls look like winners; every skip loses in half A; closed | http://159.223.88.168/hx-c9ois/sw/a1-19-pull-source.md |
| A1.20 Kit C | no Kit C passes; 3,587-byte family is a watch item (59% gate rate, 17 launches); kit edge is the tail; pull templates never reach the gate | http://159.223.88.168/hx-c9ois/sw/a1-20-kit-c.md |
| A1.21 Kit operator | hash-blind funding rule flags kits from day one (Kit B 136/140, 3,587 16/17, Kit A 79/107; 24.7% gate rate vs 0.47%); operator wallets json for the Coder | http://159.223.88.168/hx-c9ois/sw/a1-21-kit-operator.md |
| A1.22 Edge decay | flat: $ a day slope inside noise on 16 and 7 days; new buyers per trade rising (p 0.02); gates per day last 7 days slightly down (watch) | http://159.223.88.168/hx-c9ois/sw/a1-22-edge-decay.md |
| A1.23 Daily alarm | /opt/a1daily/a1daily.py at 00:15 Bangkok, appends the day row (paper + live columns), one status line, alarm on 3 fixed thresholds | http://159.223.88.168/hx-c9ois/sw/a1-23-daily-alarm.md |
| A1.10 Live test read | waiting for Ben to switch on the armed 0.0015 ETH test | (pending) |

## Asks for Supervisor
- A1.22: no size cut on decay grounds. Coder watch line: ring A1 if $1M gates per day stay at 4 or fewer for 3 days in a row. I can re-run the decay table daily with the paper book check.
- A1.21: Coder adds the `operator` tag from a1-21-operator-wallets.json and keeps the watch list learning live (kit hold stays on the hash for now). A2 or A3: money test of kit hold on hash OR operator flag vs hash only.
- No Kit C to ship. Coder / PP: tag `kit_3587` (launch `to` code 3,587 bytes, hashes 51d0e39db2ed, 2696b754c8e3, 60341fa374bf); A3 audits it for kit hold if it returns with 10+ gates.
- No gate change from A1.19; close pulls as a buyer-side signal. Optional PP counter of launches by 0x9b226f198ef0038d054be6265aa9f76c0d6e5e78 (78% instant pulls), tracking only.
- No gate change from A1.18. Coder / PP: daily health line with route mix and fresh-wallet share of post-gate buys on live-book coins.
- No size tilt goes live. Coder: log min_to_gate and gate_hour as paper tags; re-test the fixed 60-minute cut on half C (from 7 Oct) at 30+ trades.
- Close cascades as a skip rule; Coder: add a farm_funded tag (tracking only). Re-check if farm-funded coins start cascading more than about 1 in 4.
- Keep the wide gate; no early Kit B entry. With the First Window Supervisor: 83% of our kit trades are also mc5 trades, so cap combined exposure per coin and stop the two bots selling into each other.
- Coder: do NOT add the rep1_pre skip to the wide rule (it costs $302 a day); add tags kit_A, kit_B, farm_0x3393 and the Kit B bundle share for tracking only.
- PP: live counter of Kit B launches a day (bytecode hash 89885b75fd29, funder 0x84157d3d24630c297849c84d49ae7f1805c1b60f); alert if it goes to zero.
- Keep all sizing effort on v3; it is the only cell with real money a day in both halves.
- PP / Analyst 2: test v4 no-hook fee 3000 / spacing 60 entries after $500k on pp3-cross.csv, forward from 25 Sep; drop it if half B stays flat.
- Coder: the 6 A1.11 "farm" example addresses are coins, not wallets; the rep1_pre filter is unchanged.
- Coder / PP: block crews 0x04f2cb48fa0c60da3a6241077f1b7c549b9b9094, 0xa9707935dd88323fba9aa9d6a95e3ab82c1ba786 and the fee 2500 / spacing 50 class from all books; add a live guard: skip if the crew's outside sells are fewer than outside buys ÷ 10.
- PP: confirm the hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840 class definition and its quote price source.
- Coder: add `rep1_pre` to every v3 candidate (share of pre-gate buy ETH from wallets seen pre-gate in 1+ earlier gated coin); paper variant: skip if over 5%.
- Virality Expert: recompute new buyers per hour without wallets seen in earlier gated or $500k coins.

## Asks (earlier)
- Coder: paper book `w1_hot`, updated by A1.6 to A1.8:
  - Buy at 10 s when the launch tag (A1.4) names a hot crew. Use the 35 s fallback only for crews other than A and B/C.
  - Size and exit per the new final rule above.
  - Log for every trade whether the pool was pulled within 60 s of our buy. That is the live trigger chance.
- Ben, only if you choose to (no trade is placed by this session): a live test of 10 to 20 buys of 0.0015 ETH on hot A or B/C coins from a fresh wallet would measure the real trigger chance for us.
- Price Predict:
  - Free disk on main so the day-file rebuild resumes; trace the first funders of the 20 repeat wallets (list in A1.9); check the paid dRPC key (403 from main at 13:30 Bangkok).
  - Whose wallet is `0x90e2c652d8fe296dcb7fa38dff038c781fda2e23`? It made 18 early $10 buys in A and B/C on 2 to 4 Oct, and 83% were pulled.
  - Send the lab's real-mode fills per coin for the A1.7 set, so the last 10 points of gap can be split.
- Price Predict: confirm the crew.db funder row lands before 10 s live (the launch tag depends on it).


Version: v1.18 (A1.23), 6 Oct 2026 20:50 Bangkok, Analyst 1. Latest: http://159.223.88.168/hx-c9ois/sw/sw-analyst1-status.md