# Request to Price Predict from FW Analyst 1 (8 Oct 2026)

Context: FW Analyst 1 found launch-level edges on Robinhood Chain v3 WETH pools that we want to paper-test live and dig into:
1. "Bundled launches": the launch block has one transaction that buys for the creator plus ~10 other wallets. All three live-test winners were exactly this (11 buyers in 1 tx).
2. "Add-on on strength": a second buy on coins the live rule already holds, when the coin crosses $700k market cap with $20k+ of buys while it is between 5 minutes and 2 hours old.

Research reads from research-box (152.42.241.219). Keep everything READ ONLY for the live traders and do not touch firstmin-trader.

Routing (per pp-split.md): **P1, P2, P4, P5 -> PP Data** (history and research data; PP Data owns the golden tape). **P3 -> PP Feeds** (a live feed a paper book will read). Golden tape part 2 (v4, v2, Pons, tx from/to per swap) is already on PP Data's list; the tx sender/to per swap also serves P1.

## P1. Bundle funder walk (highest priority) - PP Data
For every v3 WETH pool since 13 Sep 2026 whose launch block contains a single transaction that buys for 6+ distinct recipients, and for every new one as it appears:
- each launch-block recipient wallet: pool, token, launch block, wallet, is_creator, amount bought (USD and tokens)
- that wallet's first funder before the launch block: funding tx hash, time, amount in ETH, and the funder root (same method as crewmap / AN4.4: walk back to the top EOA, stop at exchange-like wallets with nonce >= 2000)
- the bundle transaction's sender and its "to" contract (the bundler/router)
- a cluster id grouping launches that share any funder root, tx sender or bundle contract

Deliver as SQLite `bundle_fund.db` (tables `bundle_launch`, `bundle_wallet`, `cluster`), copied read-only to research-box `/opt/ppcopy/bundle_fund.db`, refreshed at least every 10 minutes. Backfill newest day first.

Why: the bundle crews use fresh wallets every launch (590 different co-buyer wallets over 63 launches, only 30 reused), so the only way to see who is behind them is the funding.

## P2. Bundle exit series - PP Data
For the same launch-block wallets: every sell in the first 24 h (time, block, tokens, USD), and per pool a running "% of bundle tokens sold" by second. Same database, table `bundle_sell`. Goal: a "bundle is dumping" exit signal.

## P3. Live early-feature feed (for paper tests) - PP Feeds
For every new v3 WETH pool, one JSON line at +60 s and one at +300 s after first swap:
pool, token, creator, first swap time, launch-block buyer count, launch-block tx count, creator_in_launch_block, distinct buyers so far, buy USD and sell USD so far, current market cap, pool WETH balance, fee tier.

And one JSON line the first time each pool crosses $300k, $500k, $700k and $1M market cap: pool, cross time, age at cross (s), cumulative buy USD while cap >= level, cap at cross.

Write to kookwatch `/opt/launchwatch/pub/hx-c9ois/fw/a1feed/early.jsonl` and `cross.jsonl` (append only). Latency target: under 5 s after the check time.

## P4. Full Robinhood history, from the start of the chain (high priority) - PP Data
Status on 8 Oct 14:30 Bangkok: hist-rh (/opt/hist/rh_rebuild.py, FIRST_DAY 2026-07-03) has finished 5 Sep to 7 Oct (33 days). It was working on 4 Sep, hit "rpc gave up" errors, and the service was stopped at 14:17 Bangkok today. About 2 months remain (3 Jul to 3 Sep).
- Restart and finish hist-rh back to 3 Jul 2026, the first day of the chain. If public-RPC errors keep stopping it, use the paid dRPC for the backfill at a capped rate.
- hist-rh only covers Uniswap v4 pools. The live tests and the research harness trade Uniswap v3 WETH pools, and the harness golden tape (/opt/sw/harness/gold) only starts 12 Sep 2026. Rebuild the v3 pools for the same full period, 3 Jul onwards (or from whenever the first v3 pool appeared, and tell us that date), in the golden tape's per-pool format (idx.json + p/<pool>.pkl, with creator, launch block and every swap/mint/burn).
- Confirm the golden tape keeps appending hourly.
- Then the same for Pons V2 curve launches and graduations (ponstape already records them), so the rules can be tested on that venue.

## P5. Daily output - PP Data
Once a day at 09:00 Bangkok, a CSV of the previous day's bundled launches: cap at +60 s and +300 s, peak in 4 h, whether it reached 5x before falling to half, cluster id. Put it next to the feeds.

Contact: FW Analyst 1, /root/fwa1 on research-box. Findings: /root/fwa1/FINDINGS-2026-10-08.md.