# 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.

DOORBELL: trig_01EZGfHD4DGnQjD6TqGqdx3X (armed, 16:37 Bangkok) [PP Feeds]
## PP Data, 8 Oct 15:15 Bangkok
- E1 runner table DONE: http://159.223.88.168/hx-c9ois/sw/e1-runners.csv (7,058 pools at $300k+). Holder columns fill from the Transfer job, then the CSV is rebuilt.
- Harness Transfers ask: running (every $300k+ token, v3 $500k coins first) into research-box /opt/sw/e1/xfer500k.db. Gold tape hourly append unchanged.
- Details: http://159.223.88.168/hx-c9ois/ppdata/ppdata-status.md

# Price Predict status (for the 2nd Window Supervisor and the first minute supervisor)

- **PP Data (7 Oct 14:05 Bangkok):** data asks now go to PP Data. Status and ETAs: http://159.223.88.168/hx-c9ois/ppdata/ppdata-status.md . E1 runner table ETA **7 Oct 22:00 Bangkok**, why: 5 of 20 shards done, 11 running on 12 workers, 4 left to start; one shard takes about 2.7 h on one worker (it waits on paid RPC answers), so the last 4 start about 17:00 and end about 20:30, plus about 1 h for holders and the CSV.

## research-box out of memory: FIXED (7 Oct 11:31 Bangkok)
- research-box RAM fixed at 11:30 Bangkok. Cause: E1 ran 20 workers on 8 cores (over the worker count <= cores rule). All 20 stopped; E1 now runs as a queue of max 6 workers in systemd unit e1run (MemoryMax 6G, Nice 10, low CPU and IO weight, flock + timeout), resumed from saved progress, not from zero. Available RAM 9.0 GB (was 141 MB). liqbuild2 untouched. E1 csv ETA moves later; exact time posted at 16:45 Bangkok.

## SW v1.15 seen (7 Oct 11:30 Bangkok)
- PP-H1 / PP-H2 not started; no handover needed. For the Harness Engineer: read-only data at research-box /opt/sw/e1/ (e1.db, x_*.db: every swap of every RH pool that reached $300k, all venues, since chain launch), /opt/hist/rh/ (v4 day files 5 Sep to 5 Oct), /opt/sw/v3hist.db + v3live.db (all v3 WETH swaps with signer), /opt/ppcopy/ (nightly copies). Plan: http://159.223.88.168/hx-c9ois/pp-plan.md


## Plan and ETAs for all asks (7 Oct 11:20 Bangkok)
- One plan for all four supervisors: http://159.223.88.168/hx-c9ois/pp-plan.md
- E1-PP e1-runners.csv ETA 17:00 Bangkok today. Golden tape v1 (PP-H1, DCPP.2/5) 8 Oct 08:00. bt.py harness (PP-H2) 8 Oct 16:00. Details and the rest in the plan.


## NEW SERVER RULES, every team (Ben's order, 7 Oct 10:49 Bangkok)

- **Read and pass to every role you run: http://159.223.88.168/hx-c9ois/server-rules.md**
- Main 159.223.88.168 is resized to 4 CPU / 8 GB (off about 10:30 to 10:46 Bangkok, back up, services checked).
- From now on research runs ONLY on research-box 152.42.241.219. Main keeps live services only.
- Never run code fetched from a public paste site (one analyst job on main did `curl dpaste | python3`). Put the script in your own folder on research-box and run it with flock + timeout + nice; exact line in the rules file.
- Please confirm in your status file which of your jobs you moved.


## Asks for Supervisor: main droplet overloaded (7 Oct 10:20 Bangkok)

- Main 159.223.88.168 has 2 CPU and 4 GB. Load average 32 to 41 at 10:15 Bangkok, swap 3.85 of 4 GB used, 30 processes waiting for CPU. Ben noticed it.
- Live trading is not on main: fmmc5 runs on firstmin 178.128.20.136 (load 0.2). But the live paper pickers and feeds on main slow down.
- Done by PP: stopped hist-rh.service (day-file rebuild, paused for disk anyway, 250 MB) and today's wash.py run (3 h 45 min). Load 21 at 10:20.
- **Ask, every team:** run research on research-box 152.42.241.219 (8 CPU, 16 GB, load 2), not on main. Main keeps only live bots and live feeds. Heavy jobs seen on main now: /opt/brain train.py, /opt/common/xsafe xsafe_tune.py (mine and canary), /opt/hist/reconcile.py every minute (300 MB), /opt/hist/feedcmp.py, /tmp/a3_xfull.py (fetched from a public paste site and run as-is), /opt/runtrack/rhledger.py (515 MB).
- PP moves its own heavy jobs (wash, crewgrade, xsafe mine, rebuild) to research-box today.


## Read this first (E1-PP ACK, 7 Oct 09:46 Bangkok)

- **E1-PP started 09:37 Bangkok.** Output: http://159.223.88.168/hx-c9ois/sw/e1-runners.csv (not up yet).
- **Pre-registered definitions (set before any result is seen):**
  - Cap = pool price x total supply x ETH/USD at that hour. Only ETH and WETH quoted pools, every venue (v4, v3 and forks, v2). Pons curve coins count from their graduated v4 pool.
  - "First time at $X" = the first moment buyers have paid $20k+ in trades at or above $X (same fair rule as pp3), so wash spikes and dust prices do not count. A raw first touch is kept in its own column.
  - Peak cap = the highest cap C where $20k+ of buys traded at or above C. The raw max is kept in its own column.
  - Runner = peak cap $3M+. Control = peak cap $300k+.
  - Split: half A = created up to 30 Sep, half B = created 1 Oct on. Full addresses, Bangkok time.
- **How:**
  - A full swap history since block 0 is about 70,000 swaps per 10,000 blocks in July, roughly 500M rows. That is too heavy for one day.
  - Step 1: a sampled scan of every pool. 1,000 blocks every 20,000 blocks (about 2 min of trading every 36 min), every venue, since block 0. Any pool trading above $300k for more than about 40 min gets caught. Faster spikes can be missed; I report how many pp3 coins (exact, 13 Sep on) the sample missed as the miss rate.
  - Step 2: an exact pass on every candidate. Full swap logs per pool, so the milestones, the peak and the 24 h / 72 h / 7 d after peak are exact.
  - Step 3: holders and new buyers at each milestone from Transfer logs, the kit hash (bytecode of the create tx target), and the wide gate pass/fail with the check that failed.
- **ETA:** step 1 about 11:00 Bangkok. Steps 1 and 2 (caps and times) about 14:00 Bangkok, published as soon as they are ready. Holders, buyers and gate columns follow the same day, in the same file.


## Read this first (v1.12 FINAL, 20:35 Bangkok)

- **pp3-series.csv.gz republished with 7 day buyers:** http://159.223.88.168/hx-c9ois/sw/pp3-series.csv.gz (md5 ca3d7990a996b3e673b11c7de4673e30, 200). Replaces the 18:24 file (md5 9972baf44632b553861149f50ef483c8, buyers only to ~48 h).
- 305,629 rows, 526 pools. Rows with buyers > 0: 28,244 (was 21,876); new_buyers > 0: 21,131.
- After 48 h (slot 289+, hourly to day 7): 81,770 rows, 7,296 with buyers, 70 pools. Check: of 71 pools with buy USD after 48 h, 70 have buyers; the 1 gap is 0xadff4ebae541e5b1077143153b77bb82f0263c868f6676b175c01edc828667a5 (not a no-log clone, left as is). 22 pools with swaps but no buyers after 48 h are sell-only.
- pp3-cross.csv unchanged (md5 7c07e3a75e1365f23e8ccc478fa291d2).

## Read this first (18:58 Bangkok)

- 7 day Transfer pass DONE 18:46 Bangkok (29,277 chunks).
- Series rerun with 7 day buyers started 18:57 (unit pp3tr7b, backup pp3.db.bak48h). The 18:24 file had buyers only up to 48 h after the crossing; hours 48 to 168 get filled now.
- ETA about 21:15 Bangkok (corrected 19:18: 78 of 526 pools in 21 min, the 7 day window reads more Transfer logs). Then republish pp3-series.csv.gz and ring the Virality Expert and the SW Supervisor.

## Read this first (v1.12 done, 18:24 Bangkok)

- **pp3-series.csv.gz republished with buyers:** http://159.223.88.168/hx-c9ois/sw/pp3-series.csv.gz (md5 9972baf44632b553861149f50ef483c8, 200).
- 304,939 rows, 526 pools, clock-fixed times (exact block anchors), one hourly ETH series.
- buyers and new_buyers now come from Transfer logs (48 h window per token, paid dRPC): 497 of 526 pools have buyers; the rest are mostly the no-log clones (nolog_clone=1 in pp3-cross.csv).
- Columns: level,pool,slot,t,cap,buy_usd,sell_usd,swaps,buyers,new_buyers,repeat_share,holders,lp_added,lp_removed_share,t_bkk
- pp3-cross.csv unchanged (md5 7c07e3a75e1365f23e8ccc478fa291d2).
- Next, automatic: the 7 day Transfer pass started 18:13 Bangkok (29,277 chunks); the series is rebuilt and republished after it.

## READ THIS FIRST (newest first, Bangkok time)
- 17:04 | **v1.12: pp3-series with the clock fix is published (swaps only):** http://159.223.88.168/hx-c9ois/sw/pp3-series.csv.gz (304,517 rows, 526 pools, every non-c10k crossing; cap, buy_usd, sell_usd, swaps, LP added and removed share, every 10 min for 48 h, then hourly to 7 days).
  - **Buyers, new_buyers, repeat_share and holders are still empty.** The Transfer download had failed. I rebuilt it on the paid dRPC: one token per call, 99,999-block chunks, 514 tokens, about 200 chunks a minute.
  - **ETA:**
    - About 18:40 Bangkok: transfers to crossing + 48 h are done, and the series is rerun with buyers (this covers your hours 4 to 8 "slow burn" test). I republish at the same link and ring you both.
    - After that: transfers out to + 7 days, then one more rerun.
  - Also fixed: the export cron raced my rebuild job (missing tmp file). It now waits on the same lock and skips while the rebuild runs.
- 16:27 | **v1.10 DONE + crossing times exact. pp3-cross.csv republished:** http://159.223.88.168/hx-c9ois/sw/pp3-cross.csv (1,076 rows: 771 at $500k, 305 at $1M; c10k 245 / 35).
  - **Times now come from the real block time.** Each block is timed by interpolating between exact chain timestamps taken every 100k blocks. Check: 0x0d4e4204903a6f0de75a30e1fd43d0c9436559f9 crosses $500k at block 65,280,844; chain time 1789639449, pp3 1789639450 (1 s apart).
    - Before: v4 times were +1 day in the first publish, and inside a day up to 74 min off. The old constant was 0.101283 s per block; the real figure is about 0.1007.
    - 117 v4 crossings sat before their own launch; now 1 does, by 1 s (rounding).
  - **The 12 coins with "no buyers around the crossing"** were caused by that +1 day shift. 0x0d4e4204903a6f0de75a30e1fd43d0c9436559f9 now crosses $500k 17 s after launch and $1M 59 s after launch, right when its Transfers happen.
  - **Second fix (found while checking):** my first rebuild silently used ETH = $2,700 on days when CoinGecko refused, so caps on 13 to 18 Sep were 7 to 10% too high. Real ETH then was $2,433 to $2,512, and this made 114 false crossings. Now one hourly ETH series covers the whole window, and the run aborts if it is missing. That is why the count went from 1,150 to 1,076.
  - **New column nolog_clone = 1:** 32 tokens (32 at $500k, 30 at $1M) are minimal proxy clones of 0xa604bed22a88ae9814476997a619c6eb876f2cc7 (code prefix 0x363d3d373d3d3d363d73). The cap_note column says how the cap was computed: pool price x launch supply, with no Transfer logs, so holders and buyers are invisible.
  - **Same clock fix in the path series and the early buyer index.** Older coins were off by hours there. The series rerun is running now (swaps only, about 4 s a coin). pp3-series.csv.gz follows at http://159.223.88.168/hx-c9ois/sw/pp3-series.csv.gz , and I ring again.
- 16:15 | **v1.11 DONE: Kit B counter live:** http://159.223.88.168/hx-c9ois/sw/kitb-count.json (per hour and per day, last 7 days, Bangkok time, refreshed every 10 min).
  - Method: every Uniswap v3 WETH pool from the v3live feed. The keccak of the create tx "to" contract's bytecode starts with 89885b75fd29 for Kit B and 0e81be52d5cf for Kit A.
  - It finds Kit B = 140 launches since 2 Oct, the same 140 as A1.14. 23 were in the last 24 h. The last was 06 Oct 13:22 Bangkok. Kit A last launched 01 Oct 15:41.
  - Funder 0x84157d3d24630c297849c84d49ae7f1805c1b60f is not used: that check needs call traces, and Kit B is a fresh kit copy per launch, so the bytecode test catches it.
  - Alarm: the JSON carries kit_b_quiet_6h and kit_a_back. My 30 min bus check rings you when either turns true (a server cannot ring a doorbell itself).
  - Self healing: a systemd timer, a research-box cron that restarts the timer, and a main cron that publishes and restarts the job when the JSON is older than 30 min.
- 15:30 | CORRECTION to pp3-cross.csv (published 14:55): for v4 rows (src "day ...") t_cross and t_cross_bkk are 1 day TOO LATE. Cause: the day files' kv t1 is the time of the LAST block of the file, and I paired it with the first block. Levels, caps, pools and tokens are right; only the time is shifted (+86,400 s). v3 / Pons rows (src v3hist.db / v3live.db) are right. Rebuilding now (~25 min); the corrected file replaces it at the same link and I ring you and the Virality Expert. TRAP for every session: /opt/hist/rh DAY.db kv t1 = time of block b_end in kv blocks "b_start-b_end".
- 14:55 | v1.9 DONE: PP3 cross table public: http://159.223.88.168/hx-c9ois/sw/pp3-cross.csv (1,150 rows = 834 coins at $500k + 316 at $1M; all columns, full addresses, plus launch_bkk, t_cross_bkk and c10k_wash_farm flag). Refreshed every 30 min. pp3-series.csv.gz appears at http://159.223.88.168/hx-c9ois/sw/pp3-series.csv.gz as soon as step 2 writes rows (raw log download for the 511 non-c10k coins is half done).
- 14:35 | PP3 step 1 DONE: fair crossing list, every venue, 13 Sep to 5 Oct: 834 coins whose sellable cap (buyers paid $20k+ at or above the level) passed $500k, 316 passed $1M. 323 of the 834 are c10k (fee 10000 / spacing 200, wash farm: kept but labelled). Table research-box /opt/sw/pp3.db cross (level, venue, pool, token, symbol, creator, fee, spacing, hooks, launch_t, t_cross, cap_at_cross, buy_usd_at_level, max_cap, src). Limit: v4 coins that were quiet in their first 30 min and ran later have no minute bars in the day files and are missed (v3, Pons graduations and the Doppler hook pools are complete).
- 14:35 | PP3 step 2 (path after each crossing: every 10 min for 48 h, then hourly to 7 days: cap, buy and sell USD, swaps, buyers, new buyers, repeat wallet share, holders, LP added and share removed) is built (/opt/sw/pp3_series.py, chain logs only, no per-tx calls). It starts when the early-buyer index finishes (~15 min), non-c10k coins first, largest first.
- 14:35 | PP6 warning: pretag was mostly noise. 1,287 of its 1,341 hop 1 to 3 tags came through shared exchange or bridge hot wallets (for example 0x88d25c861938a91af4ad57ad964a8fcc6c6351d3: 1,000 to 2,200 txs a day, first funder of creators in 16 crews). Fixed at the source: crewexport drops any creator funder seen in 3+ crews (145 wallets, http://159.223.88.168/hx-c9ois/brx/shared_funders.json); pretag cleaned and restarted. The PP6 latency report waits for a few hours of clean tags.
- 14:05 | Bus v1.2 read and followed: ring without text works (rang the Supervisor at 13:51). PP checks both supervisors' task files on every wake.
- 14:05 | v1.5 standing check DONE, daily trap table live: http://159.223.88.168/hx-c9ois/fw/pull_daily.md (per pool class AND per crew root, outside buy size $3 to $500+, share pulled within 60 s and ratio vs crew buys at the same age and price multiple; rebuilt 03:10 Bangkok nightly, republished every 20 min). Pond 29 Sep to 5 Oct: crews A and B/C pull 2x to 6x more after $10 to $50 outside buys; all pond crews pooled ~1x. c500 and c2500 small buys: 30% to 100% pulled (traps).
- 14:05 | PP1 history backfill had crashed (database lock with the live feed). Fixed: it now writes its own file research-box /opt/sw/v3hist.db (same tables), merged into v3live.db when done; restarts itself on failure. v3live current, lag 0.
- 14:05 | Day file 2026-10-05 finished on main (13:47) and copied to research-box; new day files now copy every night 02:40 Bangkok. Rebuild is now filling older days (from 9 Sep back).
- 14:05 | v1.6 signer tape: the wash rerun timer fires 15:40 Bangkok (not 14:45 as written before); signer.db goes to research-box right after.
- 14:05 | Next: PP3 fair crossing list (starts on v3live + the day files while v3hist fills), then PP2 with repeat_set.
- 13:55 | Joined the bus. Doorbell: trig_01RnKqj6VbrLa92K2pW8BFur ("Doorbell: Price Predict"). 30 min safety check scheduled.
- 13:55 | v1.7 A1.9 ask DONE: the 20 repeat wallets of crews 0x7028bab73f82dff14b199593abb76c8b6acb3aa9 (in 32 to 42 of its 42 coins) and 0x2dacdf0f44d07234832799a10a7cc5c8eff0223a are ONE group: 19 of 20 were first funded by 0x597dcbaa9edd2fc950d4f459492a223ff926b654, which buys in every coin itself and was first funded by 0x53091256ebd2d8aa37b45536a5fd864ca764f32f. Same 20 wallets in both crews. Treat them as the crews' own army. Full list with coin counts and buy ETH: research-box /opt/sw/repeat20.txt (script /opt/sw/repeat20.py).
- 13:55 | Wallet 0x90e2c652d8fe296dcb7fa38dff038c781fda2e23: real small outside bot, not crew (first ETH ~16 Sep from 0x5214ccb53636c6ebc18ce12f28dc2287b61270a5 via contract 0x97caca78ac2a94c67643d07843f85afaa44a3ea5, in no crew list, 41 coins across 5+ crews, bought 0.153 ETH, sold 0.034).
- 13:40 | v1.5/v1.7 DISK DONE: main 7.3 -> 19 GB free (two 20 Sep zst archives moved to research-box /opt/archive_main with md5 check, stale 6.5 GB liquidity WAL truncated). Day-file rebuild resumed 13:18 (2026-10-05 in progress, then 06). Disk guard on main every 15 min: under 15 GB it cleans first (journal, apt, old /tmp, old .gz, big SQLite WALs), so nothing pauses again.
- 13:40 | v1.5 dRPC: was out of balance ~12:10 to ~13:22; Ben topped it up. Feeds catch up on the paid key, then route cheap calls (logs, blocks, tx lookups) to the free public RPC; paid only for traces and archive lookups. First funder job cut to wallets that bought 0.003+ ETH (done).
- 13:40 | Crews A and B/C stopped ~09:00 (money wallets idle); new crew class fee 2500 / spacing 50 native is busy (60 launches 09:00 to 13:30, half crew-tagged); new hook 0x6404ddb4f33b4d974e570f66288e2a7ca469a840 = one fresh signer funded by an exchange-like wallet. Detail: http://159.223.88.168/hx-c9ois/fw/pp-to-sup.md
- 13:40 | v1.5 outside-buy pull test: ONE tool for every lane, research-box /opt/sw/pulltrig.py (any class, size buckets $3..$500+, pull within 5/30/60 s, matched on seconds since first trade and price multiple). Daily table per lane: next (cron + public file).

## Task status
- PP1 live v3 + Pons grad feed: LIVE, research-box /opt/sw/v3live.db (tables pools, swaps: b, t, pool, token, side, eth, usd, tok, signer, px_eth, mcap_usd, pool_eth_after, tx; liq: kind add/remove, eth, pool_eth_before, share_eth, signer). 7,551 pools. Current (lag 0). History from 13 Sep: running into /opt/sw/v3hist.db (same tables). Hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544 pools now tracked too (venue hook_4e34; history in /opt/sw/v3hook.db). Readers: ssh research-box, open read-only.
- PP6 pre-tag: LIVE, research-box /opt/sw/pretag.db (tag: addr, crew, hop, t, tx; launch: pool, b, t, signer, crew, hop, pretag_age_s). Catching up after the outage. Latency report after a few hours of clean launches (pond/c500 paused since ~10:00).
- PP7 trending boards: LIVE. research-box /opt/sw/trending.db every 3 min (DexScreener boosts top/latest and profiles; GeckoTerminal trending 5m/1h/6h/24h). GMGN trending 5m/1h + hot searches 1h every 5 min on firstmin-trader /opt/fm_pp/trending_gmgn.db. fomo: not checked yet.
- v1.6 signer tape for the Virality Expert: signer.db is filled by the wash rerun (wash, signer fill, wash) due ~14:45 Bangkok on main /opt/pricepredict-data/signer.db; a copy goes to research-box after it finishes.
- PP2 wallet score (with the new repeat_set flag), PP3 fair crossing list, PP4 X pull: NOT STARTED. Order after the feeds are current: PP3 first (supervisor v1.6), then PP2, then PP4.
- First window paper books (first minute supervisor handover): fmw1pp + fmlane1pp on firstmin-trader, paper only.

## Asks for Supervisor

## Needs Ben
- Nothing.

## Doorbell
- trig_01RnKqj6VbrLa92K2pW8BFur

Version 10, 6 Oct 2026 20:35 Bangkok. Latest: http://159.223.88.168/hx-c9ois/sw/sw-pp-status.md