# #908 Sell-probe of launcher 0x017c6044af12d5ccbfd9c4ba3639b5557be2095d (10 Oct 2026)

**Verdict: decoy kit. Add the launcher to the radar blocklist.** Nothing to buy at the pinned $2.0M price.

## Pools (all v4, USDG quote, fee 1, no hook; source /opt/sw/v4tape/usdg_pools.jsonl, usdg_liq.jsonl)
| symbol | pool_id | token |
|---|---|---|
| 中国人能飞 | 0x705d1f89ecaa9852782f2a96a40d3a19d641ff3a5a79e0d2503f422f00593aaa | 0xc20612e5db3fc93e8488b144c30f99ba5809e6e0 |
| AAPLx3L | 0xb83b9cf460196e9a85394217ce654da2097433432384da42a596937c909373e3 | 0x3f273f0bfaa4428bd7fdfb119b64be6c853f15aa |
| SI | 0x3d62b8f84044cfba42bd9a61c81bf7c92acb0cb3e4f463874874a26ec7dc4efe | 0xfdfb1083db856170404e880e741c4dc16c8e9391 |
| Muse | 0x21f3aebd80407bd6917948f85f1cdc0eb7af1805face62eabfb40130b2ab2c79 | 0xfd140122371813bb742d7a3cec04a9ac81cc6a11 |

## Method
Router eth_call sells do not work on these (see #690), so the existing sellprobed (/opt/sw/sellprobed, table probe in /opt/sw/ppdata/sellprobe.db, $20 and $100 sells per minute) was used, plus the launcher's LP add/remove history (usdg_liq.jsonl, tx_from = launcher), plus my own depth calc from the liquidity of the active range.

## Findings
1. **Liquidity is pulled.** In sellprobe.db every pool goes from ok=1 (impact ~0.0002 at a $20 sell, "reserve" $0.2-2M) to **ok=0, L=0 (impact 100%, out $0)**: 中国人能飞 within ~35 min of the first probe (by 1791631989), AAPLx3L ~1.5 h (by 1791635767), SI after ~1 h (1791637526 already out 6.7 of $20, impact 89%; 0 by 1791639466). Muse still L>0 at last probe (1791641407) but see 2.
2. **The pinned $2.0M cap sits on one-tick liquidity.** The launcher adds a 1-tick LP (ticks 338473..338474 / 483565..483566, L≈4.47e18) around the target price. USDG actually sitting in that tick = **$10.00** (wide companion range 338472 $99). The "virtual reserve" 200k-2M in sellprobe.db is the constant-L extrapolation and is false. So one $10 print moves price to the edge: that is the "$2.0M in one print", and a real sell of $20 would get $10 at the pinned price and the rest at a collapsing price (constant-L math ignores tick crossing: this overstates exit).
3. Ordering per pool (usdg_liq): add tight range at init, add far/1-tick ranges, remove tight range (~1 min), keep the far ranges for 1-1.5 h, remove all. The same 4-step pattern on all 4 pools, same funder sender 0x58daec31... = kit.
4. Not proven with a real sell (no real trade by us). Evidence is on-chain state, not simulation.

## Rule suggestions (research radar)
- Block launcher 0x017c6044af12d5ccbfd9c4ba3639b5557be2095d (tx_from).
- Generic: skip cross when pool has fee<=1 on USDG with active-range USDG < $1k (compute from L and tick width), i.e. real depth ($) must be near cap-scale, not 10 USD.
- Caveat: sellprobed "reserve_quote" overstates depth for 1-tick/thin ranges; for radar filters use the real in-range quote amount.

## Follow up: new pools from this launcher
sellprobed only tracks pools the radar sees crossing 300k. Any new pool from this launcher will show up in usdg_pools.jsonl; a watch is not built (blocklisting makes it moot). Tell me if you want a standing log.