# A1.18 Where the fresh buyers come from (Analyst 1, 6 Oct 2026, 17:10 Bangkok)

## Read this first
- **The "fresh buyers" are mostly brand-new wallets funded through two disperse contracts minutes before they buy. Nothing at the gate turns that into a filter that makes money. No gate change.**
- **(a) Routes, share of buy ETH** (all 291,899 buys in the 70 wide coins, launch to gate + 2 h):

| Route | Before gate | After gate |
|---|---|---|
| Router `0xf2a277490c6e1d511e236c64895a0fef39aa8098` | 39% | 61% |
| Uniswap SwapRouter02 `0xcaf681a66d020601342297493863e78c959e5cb2` | 28% | 38% |
| Launch kit contracts (the dev's own bundle in the launch transaction) | 31% | 0% |
| Everything else | 2% | 1% |

  - **The fomo router `0x039ec98a76f111092d4751365ff09dd2aec301e8` carried 0 buys.** I did not find a Robinhood wallet route either.
  - **Router `0xf2a277490c6e1d511e236c64895a0fef39aa8098` is a small fee wrapper** (1,617 bytes; it forwards exactInputSingle to Uniswap). Its storage slot 0 holds `0x63608970d148491dd494b2529c3cd233eb955229`, likely the owner. I don't know which app runs it.
    - **It carried 31% to 100% of buy ETH every day up to 1 Oct (81% or more on 6 of those 11 days), and 0% from 2 Oct.** The flow moved to plain SwapRouter02.
    - So one front end changed its route on 2 Oct. That is a date marker, not a coin feature.
- **(b) Funding, from 1,398 sampled buyers** (the top 6 by ETH plus 4 random per coin and phase; first ETH traced on chain):
  - **Disperse contracts dominate.** Both use the disperse selector 0xe63d38ed:
    - `0x84157d3d24630c297849c84d49ae7f1805c1b60f` (the Kit B tool): 62% of pre-gate buy ETH and 37% of post-gate in half B.
    - `0x03802649e469ed5680c58dff9e65425e22d21cc5`: 61% of post-gate in half A, 36% in half B.
  - Direct transfers from a normal wallet: 1% to 28% depending on half and phase.
  - **They are very new wallets.** Of pre-gate buy ETH, 84% (half A) and 87% (half B) came from wallets first funded less than 60 minutes before their first buy. After the gate it is 49% and 45%. The median wallet age at its first buy after the gate is 11 minutes in half B and 2 hours in half A.
  - I can't say from chain alone whether the disperse callers are trading-bot services (fresh wallet per user or trade) or farms. In the sample, 44 different callers used `0x84157d3d24630c297849c84d49ae7f1805c1b60f` and 26 used `0x03802649e469ed5680c58dff9e65425e22d21cc5`.
- **(c) Does any share at the gate split 5x trades from losers in both halves? No usable split.**
  - **Route shares can't split**: half A coins were bought almost entirely through the fee router, half B coins not at all. The mix tracks the date, not the coin.
  - **Funding via `0x03802649e469ed5680c58dff9e65425e22d21cc5`**: in half A, losers had a median 54% of pre-gate ETH from it vs 4% for winners. In half B it is 0% for both, so it can't be tested out of sample.
  - **Younger buyer wallets go with winners in both halves** (ETH-weighted pre-gate wallet age: A 2.0 h winners vs 5.1 h losers; B 0.7 h vs 1.2 h). But as a gate filter it loses money:
    - Rule picked on B, tested on A (skip if over 2.2 h): +$3,246 a day vs +$5,412 for the flat book ($1k, depth-priced).
    - Picked on A, tested on B: keeps 33 of 34 coins, +$4,221 vs +$4,142, which is noise.
    - Fixed cuts of 1, 2, 3 or 5 hours lose in both halves. The filter skips too many winners, because winners and losers overlap heavily.
- **First principles:** after the gate, our money comes from the next buyer paying more than us. Those buyers are mostly wallets created minutes earlier through public disperse contracts. They behave the same in coins that 5x and in coins that stop out. The gate already catches the moment they arrive. Who funds them doesn't predict whether they keep coming.

## Live use
- **Health signal (not a filter):** the share of post-gate buy ETH through each route and from fresh disperse-funded wallets. If fresh-wallet buying after gates falls well below the 45% to 49% seen so far, the flow our edge needs is drying up. Coder or Price Predict could log it daily from the live book's coins.
- The fee router `0xf2a277490c6e1d511e236c64895a0fef39aa8098` going quiet on 2 Oct did not hurt the rule. Half B (+$4,142 a day at $1k) kept going on SwapRouter02.

## Method
- **Buys:** token Transfers out of each pool (followed through routers up to 3 hops) from pool creation to gate + 2 h, with ETH from the pool's Swap logs. For the 57 older coins these come from A1.11's set; for 13 coins they were fetched fresh (pre-gate in A1.14, post-gate now).
- **Route** = the buy transaction's `to` (eth_getTransactionByHash for all 291,270 transactions). Launch kit = any `to` with the Kit A, Kit B or 3,587-byte kit bytecode (A1.14).
- **Funding:** binary search on eth_getBalance for the first block with ETH, then that block's transactions (direct transfer, or a call whose input names the wallet). Wallet age = first buy block minus first funding block.
- **Money:** Analyst 2's wide trades at $500 (x with the 0.3x stop at block end), re-priced to $1k on each pool's WETH at the gate (as in A1.17). Halves: A = 21 to 27 Sep, B = 28 Sep to 6 Oct.

## Limits
- Funding is from a sample (about 20 wallets per coin, 1,398 in all), weighted by ETH.
- Bonus (the 490 instant-pull coins) not done in this pass.

## Asks for Supervisor
- No gate change.
- Coder / PP: add a daily health line with the route mix and the fresh-wallet share of post-gate buys on live-book coins.
- If anyone can name the app behind fee router `0xf2a277490c6e1d511e236c64895a0fef39aa8098` (owner slot `0x63608970d148491dd494b2529c3cd233eb955229`), that tells us which front end brought half A's buyers.

First principles score: 80%. Routes come from every buy transaction, and funding from each sampled wallet's first-ETH block on chain. Missing: full buyer coverage, an outside label for the disperse callers (service vs farm), and the instant-pull comparison.

Scripts: /root/a1/a118/post13.py, buys.py, txto.py, routes.py, split.py, fund.py (review container). Latest: http://159.223.88.168/hx-c9ois/sw/a1-18-buyer-source.md