---
name: delta-liquidity-2
description: Continuation of delta-liquidity.md from 6 Sep 2026, after that file hit its size cap. The TVL-is-unverifiable correction, the usage-vs-price update, the DELTA alert bot, and the deployer-graph / no-hook re-verification. Read both whenever DELTA comes up.
sources: [cowork]
aliases: [delta-2, deltaliquidity-2, delta-alert]
---

Continues [[delta-liquidity]]. Same token 0xe8ffd7e24187F72afB08d75B1bb13088A989a791, Robinhood Chain (4663).

## 6 Sep 2026 ~07:00 UTC — TVL IS NO LONGER VERIFIABLE, and the watcher is live
The 7702 treasury delegation and the treasury selling its fee tokens are already recorded in [[delta-liquidity]] and are not repeated here.

### NEW CORRECTION — stop quoting the app's TVL as a measured fact
- Delta's app claims **$2,007,718** (tvlUsd in the RSC payload). Re-running the exact 31 Aug on-chain method verifies only **$23,371**.
- The custodied v3 book **collapsed from 473 positions / $381,637 (31 Aug) to 120 positions / $23,371**, while total positions grew to 13,965. All the growth moved to Uniswap **v4 (10,147 of 13,965 positions)**, which lives in the PoolManager singleton and cannot be reached by NFPM enumeration. Searched Delta's bytecode for a v4 position manager: **not found**.
- **MC/TVL for DELTA is therefore unverifiable. Never quote 5.7x or any TVL ratio as measured.** This voids the "MC/TVL 19.4x, mid-band" line in the 31 Aug block of [[delta-liquidity]]. **Revenue is the only number that carries the case.**

### Usage vs price, the update Ben asked for
| | 31 Aug | 6 Sep | change |
|---|---|---|---|
| Positions | 5,158 | **13,965** | +171% |
| of which Uniswap v4 | 2,540 | **10,147** | +299% |
| Lifetime fees paid to LPs | $508,068 | **$2,303,522** | +353% |
| Holders | 12,740 | **15,364** | +21% |
| Protocol revenue, measured | $244,000/yr | **$1,131,604/yr** | +364% |
| Market cap | $29,275,580 | **$11,010,990** | -62% |
| **cap / revenue** | **120x** | **9.7x** | 12x cheaper |
- Revenue agreed within 8% across two independent methods: the treasury log scan over an **exact 7.03-day window derived from real block timestamps** (Robinhood Chain measured at 9.92 blocks/sec, but nothing relies on that), and 1% of the change in Delta's own WETH fee counter.
- Chain context: TVL $908,649,407 (+108.8% 30d, +29.6% 7d), DEX volume 24h $1,612,673,902, chain fees 30d $188,156,495 (+422% 1m).
- Chart, against Ben's "never call it late at -60/70%" rule: **RSI 51.9, not oversold**, holders still rising, volume collapsed from $11.1M (30 Aug) to under $1M. Reaccumulation shape, but no bid and no dated catalyst.

### VERDICT 6 Sep: HOLD
Cheap on real revenue. The token receives none of it. `feeBps()` = 100, all three fee contracts pay the same treasury, no payout code exists anywhere. 100% of the price is an option on one decision by 0xb1c2bbf86e557ecdc1812f75ae3fe973e5ec9e69.
Ladder at PONS's 4.72x cap/revenue: route 100% at today's 1% take = **$5.6M, BELOW spot**, so a route alone is not enough. 5% take = $27.9M. 10% = $55.8M. Snuggle's 14% = $78.2M. **The bull case needs the route AND a 5x-14x take-rate rise, together.**

## "DELTA alert" — the bot, built 6 Sep 2026. Full build detail in [[alert-bot-fleet-2]].
`deltaalert.service` at `/opt/delta-alert` on the kookwatch droplet, live, enabled, in botctl (14 bots) and watchdog (13 units). Delivers to @bk_rhwatch_bot, chat 357464289.
- **Fee-switch watcher, every 60 seconds.** Fires once, loudly, on any change to: `feeBps()` on 0x5cA6214227D1195c4b7b4B96847b8966c688295D, `treasury()` on all three fee-bearing contracts, `owner()`, the treasury's CODE (a real payout contract is the loudest case), DELTA arriving at 0x000000000000000000000000000000000000dEaD, or the treasury accumulating DELTA. Baseline seeded at today's values so it only speaks on change.
- **Daily card at 01:00 UTC = 08:00 Asia/Bangkok**, collected **three times, median used, and any field whose three passes disagree beyond a per-field tolerance is NAMED in the message** rather than silently averaged. Plus a fourth independent check: revenue recomputed as 1% of the fee-counter delta since yesterday's card, flagged if the two methods differ by more than 35%. If all passes fail it sends "report FAILED" and no numbers.
- First live run 6 Sep 06:55 UTC: 3 passes in 30s, 25s, 23s, **zero flags**.
- Leads with a plain action word (STRONG BUY / BUY / HOLD / TRIM / SELL) per Ben's 2 Sep doctrine. It is not a repeating status board: the fee-switch half only speaks on change.
- **Ben no longer has to ask for this update. The triggers it watches are exactly the ones named as what flips the verdict.**

---

# 6 Sep 2026, later — FULL RE-VERIFICATION PASS. Ben: "check everything you said again." FOUR of Claude's own claims were wrong.
- [stated] Ben asked specifically what the fee-wallet activity really means, what else could explain it, whether the team or anyone else made a statement about it, and for more detail on the "AMM revenue comes later" claim.

## ERROR 1, the worst: "zero contracts deployed, nobody is building in public" was FALSE. The wrong wallets were checked.
**There is a THIRD team wallet and it is the real deployer: 0xf98c1097bc50692b8f290b8c20a3f3f7dd0e2a1d.** Found by calling `owner()` on the factories: VaultFactory 0x68EDc4948F60D21c4a7Dcbb8Ed4500cE6D0b153c and VaultFarmFactory 0x2bdA3FeB985d812a5932fe59eD4D8627BA3A10d1 both return it. LadderManager 0x5cA6214227D1195c4b7b4B96847b8966c688295D returns 0xb1c2bbf86e557ecdc1812f75ae3fe973e5ec9e69 instead, which is why only the keeper wallet was found.
**31 contracts deployed, weekly cadence, ACCELERATING:** 21 Jul (3), 29 Jul (16, the original stack), 3 Aug 0x6235cf6bd8419b34942f4eddb39c880bd96dd700, **15 Aug 0xc5941433114bb47a9733cb31a0a3a3dbff45b418, 23 Aug 0x5cA6214227D1195c4b7b4B96847b8966c688295D, 28 Aug 0xac6a35d097bb230eb9966ddd8e5970055a622d98, 29 Aug 0x077af3f17c0ef7bb04a9882e1955e9c23d2f9e91, 1 Sep 0xf321ed71750c3e627a0f435c8950429f231e3497 + 0x95d0db9c60a50234103ce1a387e43b2fddd262b2, 2 Sep 0x56bf0f5b26d888b989a9a20ec2164f0ca560836c (47,092 bytes of init code, their largest ever).**
**STANDING METHOD FIX, applies to every project: never conclude "not shipping" from the keeper wallet. Call owner() on EVERY factory and manager first, enumerate all owner wallets, then check CREATEs on each.**

## The hook/AMM answer got STRONGER, not weaker. They ship weekly and they ship MORE OF THE SAME LAYER.
Probed the runtime of all six recent managers:
- Identical selector set every time (0x909b19d9 keeper, 0x5b1b953e, 0xffb94054, 0xd51e9064, 0xf2c4fa40, owner()), same owner 0xb1c2bbf86e557ecdc1812f75ae3fe973e5ec9e69, same `treasury()` 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f. Runtime grows 18,438 to 22,360 bytes. **Same product, versioned.**
- **Every one references the V3 NonfungiblePositionManager 0x73991a25c818bf1f1128deaab1492d45638de0d3 and WETH 0x0bd7d308f8e1639fab988df18a8011f41eacad73. NOT ONE references the Uniswap V4 PoolManager 0x8366a39cc670b4001a1121b8f6a443a643e40951.**
- **MECHANICAL PROOF NO HOOK IS COMING: every deployment is plain CREATE from an EOA (sequential nonce). A Uniswap v4 hook address must encode its permission bits in the low address bits, which requires a CREATE2 address search. This wallet has NEVER used CREATE2. No hook is in preparation. Reusable test on any chain.**
- **`feeBps()` (0x24a9d853) returns 100 on ALL SIX, including the one deployed 2 Sep. Six fresh contracts, six chances to raise the take rate, still 1.00% every time.** `MAX_FEE()` (0xbc063e1a) reverts on all.
- **ESTIMATE GIVEN TO BEN: v4 hook inside 6 months ~15%. Own AMM or venue ~10-15%.** (An earlier answer the same day said 35-45% / 15-25% and was revised down on this evidence.)

## ERROR 2: "the treasury cashed out the fees" was an OVER-READ. Corrected by decoding every leg.
Decoded all Transfer logs on the 6 Sep 06:18-06:28 burst. **They sold ~25 memecoin fee positions (ROBINCAT, microduck, QUOTRON, NEST, $ROBBIE, Rabbit, DTF, BONER, HOOKR, VISTA, AMC, RAM, DINO, COPPERINU, INDA, CINEMA, MOO, GRASS, NUDES, PARE, GG, INJOH, PAIR, Index) and received WETH on every single leg, ~1.38 WETH in that burst.**
- **The WETH DID NOT LEAVE. Treasury holds 4.043114 WETH (~$10,160 at ETH $2,513) plus 0.496634 native ETH. Nothing bridged, nothing sent to an exchange.** This is CONSOLIDATION of dust into one asset, not an exit. **Any line in these files saying the team "sold the fees rather than buying the token" must carry that qualifier: they converted, they did not withdraw.**
- Zero DELTA bought. Treasury DELTA 67,136.84 = ~$737 = fee dust from DELTA-paired LPs.
- **The notable part is the TIMING: the treasury has 38 transactions in its ENTIRE LIFE and 30 of them were on 6 Sep.** Prior: 3 on 21 Jul, 2 on 29 Jul, 1 on 29 Aug, 2 on 1 Sep. **First real consolidation the treasury has ever done.**
- **Three readings, not yet separable: (a) housekeeping finally done because 274 dust tokens got unwieldy, (b) consolidating into one asset BEFORE doing something with it (seed liquidity, a buyback, paying a bill), (c) preparing to move it out. THE TELL: watch whether WETH LEAVES 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f. If it leaves, (c). If it grows and sits, (a) or (b).** The deltaalert bot already watches this address.
- Treasury identity RE-CONFIRMED independently: `treasury()` (0x61d027b3) on all three fee-bearing contracts returns 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f.

## ERROR 3: the "AMM revenue comes later" quote is MISATTRIBUTED in [[delta-liquidity]]. It is not the founder.
Read the actual post: **https://x.com/MCGlive/status/2088713881783074882, dated 16 Aug 2026 (the older file says 11 Aug), 39.3K views.** It is **@DineroDom0's thesis relayed by @MCGlive, not the founder speaking.** Content is the Meteora analogy: Meteora captured fees regardless of where tokens launched, 80% of revenue from non-Meteora pairings, ~$120M profit last year after distributing 90% to LPs, "nobody has filled that role on Robinhood Chain yet."
**@omen_xbt's ONLY words on that thread: "great synopsis, we're just getting started."** @deltaliquidity's only reply: "Deltamode."
**So the AMM commitment credited to the founder is NOT a verified founder statement. It is a promoter's framing the founder liked. That removes most of the evidence base for the AMM thesis.**
Two new items from that post, both CLAIMS not facts: (1) **"The team brings a DLMM background from Solana with proprietary technology already built"** — against the founder's own admission that his prior Solana project was bootstrapliq, a retired pump.fun hackathon project, this is generous and unverified. (2) **A named competitor not previously on file: DLMM is competing for the same slot on Robinhood Chain.**
@omen_xbt as of 6 Sep: **875 followers** (was 864), 251 following, joined February 2026, bio "@deltaliquidity onchain".

## ERROR 4 / measurement trap: the app's "Total Fees" counter is MARK-TO-MARKET, not cumulative.
It read $1,624,254 on 4 Sep and **$1,612,545 on 6 Sep from the docs-page header. A cumulative USD counter cannot fall.** It revalues accrued fees at current token prices. **Never use the app's Total Fees as a cumulative revenue time series; it falls in a drawdown even when fee accrual is positive.** Any growth rate computed from two readings of it is invalid. (The separate $2,303,522 figure in the table above came from a different, larger-window read and the two should be reconciled before either is quoted again.)

## DOCS re-read in full (deltaliquidity.app/docs, single page, 71,705 bytes)
**Zero occurrences of: AMM, hook, roadmap, buyback, burn, governance, revenue share, $DELTA, or the token in any form.** "V4" appears once and only to say Delta *finds* V4 pools alongside V3 and USDG pools and routes to the deepest. **The docs describe a reader of other people's pools, not an owner of pools. Nothing published commits them to becoming a venue.** deltaliquidity.xyz exists but returns empty; live docs are deltaliquidity.app/docs.

## Open item for Ben, blocking X work across the whole fleet
**twitterapi.io is out of credits** (`recharge_credits: -146`, HTTP 402). Any X query returning zero right now means the key is dry, not that the account is quiet. Recharging is a step only he can do.
**Workaround found and verified: X SEARCH is login-walled in the built-in browser pane, but INDIVIDUAL TWEET PERMALINKS and profile headers render fine logged out and can be read with Claude_Browser__get_page_text.** Claude in Chrome returned "No tab available" this session.

## 6 Sep 2026, later still — "DELTA has no revenue" was written to Ben. He caught it. THE ERROR AND THE GATE.
- [stated] Ben, on a bottoming screen that reported DELTA as having no revenue: "All your information is wrong. For example, the DELTA no revenue is wrong."
- **THE ERROR: I read DefiLlama's Robinhood Chain fee list (150 protocols), found no row containing "delta", and wrote "no revenue line on DefiLlama at all" as a fact about the token. Both DELTA files already said, repeatedly, that DEFILLAMA DOES NOT TRACK DELTA'S FEES. Absence from an aggregator is never evidence of absence. Same trap shape as the fomo-holder-absence trap that has now misled three times (BUNNY, Shoppy, this).**
- **STANDING GATE: never write "no revenue" about any project until the treasury measurement has been RUN. Script kept at /tmp/bt/deltarev.py; it takes 4 minutes. If it cannot be run, the honest output is "not measured this pass", never "no revenue".**

### Third independent revenue read, and it CONFIRMS the 9.7x block above
Same method (chunked `eth_getLogs` Transfer, topic2 = treasury 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f, no address filter, price every token via DexScreener `tokens/v1/robinhood` 30 at a time), 14-day window:
- **1,750 inbound legs, 477 distinct tokens, 465 priced, $26,458 total. Last 7 days $25,226 = annualised $1,315,337.** The block above measured $1,131,604/yr on a 7.03-day window. **Two independent windows, 16% apart. The revenue is real and the order of magnitude is settled.**
- Cap $10,959,074 at price $0.01095 gives **cap/revenue 8.3x**, against 9.7x in the earlier read the same day. Series: **120x (31 Aug) -> 36.5x/15.5x (4 Sep) -> 9.7x -> 8.3x (6 Sep).**
- Batching re-confirmed, never read a 24h window as the rate: 31 Aug $618, 1 Sep $5,870, 2 Sep $5,177, **3 Sep $3.82, 4 Sep $2.58**, 5 Sep $9,544, 6 Sep $3,502.
- Same-chain annualised revenue for scale: Pons V2 $344,947,108 · up v3 $35,563,514 · PAIR $6,175,330 · Snuggle $5,747,133 · Alandale/LUTE $2,711,167 · **DELTA $1,315,337**.

### Price shape 6 Sep, measured on the DELTA/WETH base pool 0xd64fbda67e1015df43fa5e49f02ca844729e5f94
ATH 6.92 days old, **-70.1% off it**, the post-ATH low printed in the CURRENT hour, 7d **-61.7%**, 24h **-26.4%**, vol7d $44,731,544 = **2.53x the prior week**. Falling on RISING volume with the low still being made. **No base, and this contradicts the "volume collapsed to under $1M" line in the earlier 6 Sep block: that was one pool, this is all base pairs. Use the all-pairs figure.**
Launch cohort re-run at 3 hops with contracts filtered at hop 0: only **5 wallets** (0xe8b41133fe0f86526bf33df974bc26a63b608440, 0x79e1a2f6254619c5c29971913c2188b44124cfbb, 0x1978df99eb6fca286455c6fc279cd98582f1a8d5, 0x399fae9b81c1a4e447736d608e6e1c03ad7472c3, 0x743a5abb39d5eee7257685f5c9938b12e9d7df0e), sold 5.88% of supply, **all hold $0**. **There is no insider overhang left to blame. The decline is demand, which is harder to fix than an overhang.**

## 7 Sep 2026 22:15 UTC — FIRST BUYBACK AND BURN. The fee switch fired, unannounced. Verified on chain, 8 Sep.
- [stated] Ben pasted the deltaalert "DELTA IS BEING BURNED, +2,414,692 DELTA, this is the trigger" card and asked "is this true?"
- TRUE, and the alert under-described it. Sequence, all from treasury 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f: 20:49:49 UTC unwrapped 6.332 WETH to ETH (tx 0x0489a38cd1...). 22:14:53 sent **9.96 native ETH (~$25,000) to the universal router 0x8876789976decbfcbbbe364623c63652db8c0904 and received 2,339,169 DELTA** (tx 0x436bdb4cb20fee8c57aeb9bd3433a45fe77f9930d473bffb8563984d52186b52). 22:15:41, 48 seconds later, **sent its entire DELTA balance, 2,414,691.72, to 0x000000000000000000000000000000000000dEaD** (tx 0xe4170258ab03eb16f5cf7d6684b4ff746de9acbf87c1ad6c4e0d836208249776). Treasury now holds 0 DELTA, 0.0000179 WETH, 0.0017 ETH: emptied
- Total DELTA ever burned 2,414,745 = 0.24% of the 1B supply. Bought at about $0.0107, roughly 1.1% of the 7-day traded float
- The 6 Sep "three readings" question is answered: reading (b). The dust consolidation on 6 Sep was the treasury gathering ~a week of fee revenue into one asset to buy and burn the token. Nothing left the treasury to an exchange
- What it is and is not: $25,000 is about ONE WEEK of measured revenue ($1.13M to $1.32M/yr), so this was a 100% route of a week's fees at the unchanged 1% take. No contract changed: feeBps still 100, no payout code, no announcement found. It is a discretionary keeper action by the same wallet, not a mechanism. One burn is an event; the second one makes it a program
- Market reaction by 8 Sep 00:30 UTC: price $0.01354, cap $13,508,782, +1.7% 24h. Barely noticed
- The alert bot's line "TREASURY IS ACCUMULATING DELTA" that preceded it was the 2.34M buy landing at 22:14; it fired correctly, one minute before the burn

## 7 Sep 2026 22:14 UTC — OFFICIAL TOKENOMICS PUBLISHED. The two conditions the 6 Sep verdict demanded both landed in one night. Verdict flips HOLD -> BUY.
- [stated] Ben pasted x.com/deltaliquidity/status/2097086136917348363 (an X Article "$DELTA Tokenomics", 28.6K views, 158 likes, posted 22:14:31 UTC, the same minute as the buyback)
- WHAT IT SAYS: protocol fee rises from 1% to **7.5% "effective today"**; revenue split **80% weekly buyback-and-burn / 20% operations**, drifting 5 points a month to 60/40 by Jan 2027; no fixed amount, scales with revenue; their own figure: users claimed ~$2,000,000 of LP fees in the last 7 days, so 7.5% = $150k/week = $600k/month; supply 900M public, **100M locked until Jan 2027** with disclosure before any move; "parameters subject to change"
- VERIFIED ON CHAIN: **new manager 0xbCb96b15dC2246D242c879316c86e25e846ad5FB, deployed 21:23 UTC 7 Sep by 0xf98c1097bc50692b8f290b8c20a3f3f7dd0e2a1d (51 minutes before the article), returns feeBps() = 750**, treasury 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f, owner 0xb1c2bbf86e557ecdc1812f75ae3fe973e5ec9e69, 22,360 bytes, references the V3 NFPM and NOT the v4 PoolManager (same product line, no hook). **All six older managers still return feeBps 100**, so positions already open on them keep paying 1% until they roll to the new one. The first burn (2,414,692 DELTA, ~$25k, block above) executed 48 seconds after the article
- Their "$2M user fees in 7 days" agrees with my independent measurement: $25,226 of treasury inflow in 7 days at 1% implies $2.5M of user fees. The base is real
- NOT VERIFIED THIS PASS: the 100M lock (Blockscout holders endpoint down). Treat as claimed until the lock contract is read
- REVALUATION: measured base revenue $1.13M to $1.32M/yr at 1% becomes **$8.5M to $9.9M/yr at 7.5%** if usage holds (their own number: $7.8M/yr). 80% burned = **$6M to $8M/yr of buying against a $13.5M cap, about half the market cap per year**. At PONS's 4.72x cap/revenue the ladder reads **$37M to $47M, 2.7x to 3.5x from $13.5M** (the 6 Sep ladder had 5% = $27.9M and 10% = $55.8M, so 7.5% ≈ $42M, consistent)
- RISKS that keep it from STRONG BUY: a 7.5x fee raise can shrink LP usage; the burn is a weekly keeper action, not contract-enforced; "subject to change"; old managers still at 1%; 100M unlock Jan 2027 = 10% of supply
- The 6 Sep sentence "the bull case needs the route AND a 5x-14x take-rate rise, together" is now satisfied on both halves. Market reaction so far: +1.7% 24h. Not priced

## 8 Sep 2026 00:48 UTC — SELLER LEDGER (venue-only sells, full history from launch block 24,483,783). Price $0.01312, cap $13,097,667
- [stated] Ben: "Are there any insider sellers left in DELTA? Any of the bundlers? Any huge wallets keep selling?"
- METHOD: Dune is over its datapoint cap and Blockscout was down, so the seller_ledger logic was re-implemented on the RH RPC: 1,640 chunks of 20k blocks, 711,785 outbound legs, 39,328 addresses, zero dropped chunks (/tmp/fs/delta_ledger.py, aggregates in memory, never stores the full leg list). Venue set = 105 auto-derived (>300 legs, net-flow ~0, balance <1000) + 5 DexScreener pools + known routers = 113. Two more pass-through contracts surfaced as "sellers" and must be added to the venue set next run: 0x22ade85e03f89cf8465287d726734ffde6d3f891 (20,392 bytes, 7,832 legs, in 184.7M out 184.4M) and 0x48a097df16c7844a33b1c3d11ab353457846e13f (2,650 bytes, owner 0x2bfd1fc5e25a8f55c2e849492ad7966ea8a0dd9e, in 162.8M out 162.2M). Cost basis was NOT computed this pass (gate 5 open)
- INSIDERS: NONE. Deployer 0xf98c1097bc50692b8f290b8c20a3f3f7dd0e2a1d and keeper 0xb1c2bbf86e557ecdc1812f75ae3fe973e5ec9e69 have never held a single DELTA. Treasury only passed the 2.41M it burned. The five launch-cohort wallets are 100% out (net 0 each). **The 100M "locked" supply is real as a holding: 0xd0f7d8c6e9f6d80c297bebe4f7fd1b9c8125c32f, a 4,941-byte contract, 100M in, 0 out, the same locker type that holds 22.6M PARE.** Unlock date not read from the contract yet
- ONE BIG ACTIVE SELLER: 0x551ee8420d4a1711e5f63eb47c532738940805c8 (7702), acquired 45.2M, sold 28.75M (64%), 6.35M in the last 24h, 2.9M in the last 6h, still holds 16,476,124 (~$216k, 1.65% of supply). Last sell 7 Sep 22:03. Everything else selling in the 24h window is under 1M left or a router
- BIG HOLDERS SITTING STILL: 0xda4bcee76b29efec9697fcf663601c2042043968 EOA 29,409,646 (2.94%, never sold a token); 0x2dbaf98620ac5bbe3441f756feff82702d095b1a 23,742,490 (the exchange hot wallet); 0x1b72ee84b5675809d3b061a0046798cd61b31de1 20,780,990 (sold 13%, last 7 Sep 12:54); 0x736274051aed34ddd62ade04417bca12a06148af 19,717,752 (sold 56%, silent since 1 Sep); unipcs 0x0a6ebed0155edb4b21d92ad02897a626cd90119e 15,980,267 (sold 45%, silent since 4 Sep); 0xc623604ecd3b9014768c61db4bebf37f77a5ab3b 15,754,368 (sold 16%); 0xe7d64660e59bbd7651c1e31f45a7fc5d30b981db 11,500,000; 0xe48b6e0e9900d68ba93ac417d9e2ce911c52f753 10,993,547; 0x7e3ba68c49561aae7c23c1d20fef0f1d7615a3ad 8,891,469 (the rank-1 copytrade wallet, sold 42%)
- OVERHANG: 116 wallets are >=50% out and still hold 50k+: 33 active ($436,717) + 83 paused ($1,134,782) = 119.8M DELTA = $1.57M = 12% of cap. 2,158 wallets have finished (dumped 14.9bn of churn, hold 328k)
- 24h venue sells by >1M sellers: 123.3M DELTA ($1.62M) gross, about 88M ($1.15M) after removing the pass-through contracts. Volume-driven churn from 7702 fomo wallets flipping 100% in and out (0x3f2fd492c623d6800ea35830651ea8206c8ebc14, 0xba981d436d0e0f76925228581447dfcb7cd3ad08, 0x135413bb6883d648c4faf261e7d3a9e6ede740a1, 0x397abdfd1b850d6baaece47e5dc7c826a0f246c2 each bought and sold their whole stack inside the day)

## 8 Sep 2026 01:00 UTC — THE DAILY CARD WAS WRONG TWICE. Ben caught it. Bot patched.
- [stated] Ben pasted the 8 Sep daily card ("BUY, cheap on real revenue, but the token still gets none of it, $8,298/day, revenue +126%, the two methods disagree by 82%") and asked what it means, why the alert sends it if false, and to fix the methodology
- DEFECT 1: the revenue scan counted EVERY token arriving at treasury 0x266311a44e70363b3d168bcb38cf38cbf2ccc51f as revenue, so the 2,339,169 DELTA the treasury BOUGHT BACK (~$25k, from router 0x8876789976decbfcbbbe364623c63652db8c0904) was booked as income. That doubled the number (+126%) and is the whole 82% disagreement with the fee-counter method. RULE, applies to every treasury bot: a treasury inflow from a router or a pool is the treasury's own money moving, never revenue. Revenue = legs whose `from` is a fee manager contract (treasury()==treasury), nothing else
- DEFECT 2: feeBps was read only from the 23 Aug ladder 0x5cA6214227D1195c4b7b4B96847b8966c688295D (still 100), so the 750 on the new 7 Sep manager 0xbCb96b15dC2246D242c879316c86e25e846ad5FB was invisible and the verdict text "token still gets none of it" stayed on after the buyback and burn
- FIX (deltaalert.py on kookwatch, backups deltaalert.py.bak-* and config.json.bak-*): config now lists deployer 0xf98c1097bc50692b8f290b8c20a3f3f7dd0e2a1d, all 7 managers, burn_share_pct 80. Bot derives CREATE addresses from the deployer nonce (pure-python keccak, RLP [sender,nonce]) and probes each for feeBps/treasury, so a new fee contract is found and alerted automatically ("NEW DELTA FEE CONTRACT DEPLOYED"). feeBps reported per manager; verdict uses feeBps_max; route_live = burned>0 or feeBps_max!=100 or treasury has code; card now says "Goes to holders ~$X/yr via weekly buyback and burn (80% share)" and "Take rate 7.50% on new positions (N contracts), 1.00% still on M older ones" and "Burned so far X DELTA"
- TEST RESULT 8 Sep 01:45 UTC after three follow-up patches: revenue $20,215/7d = $1,055,285/yr (buyback's 223 legs excluded), feeBps 750 on the newest manager, feeBps_max 750, 16 contracts still at 100 (6 managers + 10 per-user vault clones of 10,505 bytes that also sweep to the treasury and answer treasury()==TREAS), burned 2,414,745, cap/revenue 12.8x, verdict route_live -> STRONG BUY. Service restarted, active; corrected card sent 01:44 UTC
- THREE MORE BOT TRAPS fixed the same hour, reusable on every bot: (1) a generic rpc() that retries on "execution reverted" turned one probe per pool into 18s of backoff, so the daily scan ran 9+ minutes with no output: a revert is an answer, never retry it; (2) a transport failure was cached as "not a manager" and poisoned the cache, so every leg was skipped and revenue read $0: only cache what the RPC actually answered; (3) the fee-counter cross-check assumed one take rate, but 1% and 7.5% are live together, so it now checks against a band [1% x counter, feeBps_max x counter] and skips windows under 12h

## 13 Sep 2026 — THE BUYBACK AND BURN IS A PROGRAM ON A 3-DAY CLOCK. Full burn history, verified on chain.
- [stated] Ben pasted the "TREASURY IS ACCUMULATING DELTA" + "DELTA IS BEING BURNED" pair and asked "is this new? tokens didn't get burned before? treasury didn't accumulate delta before?"
- **ANSWER: not new, it is the third. Both alerts describe ONE event 25 seconds apart.**
- **EVERY burn to `0x000000000000000000000000000000000000dead` in DELTA's life, by `eth_getLogs` topic2=dead over the full chain history:**

| when (UTC) | DELTA burned | ETH spent on the buy | buy tx |
|---|---|---|---|
| 07 Sep 22:15:41 | 2,414,692 | 9.96 | 0x436bdb4cb20fee8c57aeb9bd3433a45fe77f9930d473bffb8563984d52186b52 |
| 10 Sep 21:59:08 | 912,097 | 4.24 | 0x1992e3ee28041a3fa77441aca301482d71c4d6a356d6d47433d80acfc027823b |
| 13 Sep 01:14:34 | 993,449 | 4.136 | 0x22ff83b4e28149b05c9235fe8a003e6bf57604ddaa29862e44681a824596e5af |

- **Total 4,320,291 DELTA = 0.432% of supply, $45,925 at $0.01063.** Matches the bot's counter exactly. Burn txs `0xe4170258ab03eb16f5cf7d6684b4ff746de9acbf87c1ad6c4e0d836208249776`, `0xcde52035af7eeff0165a4d2f1d6bdf19781b51c93133328246b547dacf5ea7d1`, `0x746c581214322de764657d9f95e8c2fa3227137c53c1a39d426262486d451862`.
- **Before 7 Sep the burn address had received 53 DELTA total**, in three dust legs from `0x7764b09bcf9fbb9140c02bd65e38e28bb956e51d` (13 Aug, 0) and `0x5d754384a7c0503dbe1b67595aca7ebe1dad305e` (2 Sep, 27 and 26). Nothing from the team. So the program genuinely starts 7 Sep.
- **CADENCE IS EXACTLY 3 DAYS: 7th, 10th, 13th.** The tokenomics article announced "weekly". They are running it **more than twice as often as announced**. Three consecutive on a clean clock is the strongest evidence yet that it is a real program rather than a one-off gesture.
- **SIZE SETTLED AFTER THE FIRST ONE.** Burn 1 was 9.96 ETH, a catch-up of the backlog. Burns 2 and 3 were 4.24 and 4.136 ETH. **Steady state ~$10,500 every 3 days = ~$3,500/day = ~$1.3M/yr of buying against a $10.59M cap, about 12% of market cap per year.** That is consistent with the measured 7.5% take on ~$1.3M/yr of base revenue.
- **THE ROUTER CHANGED and no bot should key off it.** Burn 1 routed through `0x8876789976decbfcbbbe364623c63652db8c0904` (24,546 bytes). Burns 2 and 3 through `0x0000000000001ff3684f28c67538d4d072c22734`. The DELTA that landed in the treasury on burns 2 and 3 came from `0xaa61254627b7392b0bc922097b10eb0587db2be7` (17,562 bytes).
- **Treasury DELTA inflows split cleanly into two classes and the threshold between them is wide:** fee revenue paid in DELTA arrives from the fee managers in legs of **18 to 14,000 DELTA**; the buybacks arrive in legs of **900,000 to 2,340,000**. A 100,000 floor separates them with an order of magnitude of headroom either side. Treasury DELTA outflows: **three, ever, all to the burn address.** It has never sent DELTA anywhere else.
- Treasury native ETH after burn 3: **0.0168 ETH.** Emptied every cycle, same as after burn 1.
- Price is not reacting: **$0.01063, cap $10,587,717, -1.5% 24h**, 20 pairs, agg liq $1,103,212, agg v24 $1,634,958. Three burns in and the market has not repriced it.
- STILL TRUE and still the ceiling on the thesis: it is a keeper pressing a button. `feeBps` is set, but nothing in any contract compels the buy or the burn. It can stop on any cycle with no warning, which is exactly what the bot now watches for.

## 17 Sep 2026 09:15 BKK — "is this a new fee?" NO. TWO REPEAT ALERTS AND TWO FAKE BURN ALERTS, ONE ROOT CAUSE. BOT PATCHED.
- [stated] Ben pasted two "DELTA alert — FEE SWITCH / NEW FEE CONTRACT SEEN ... feeBps = 750 (7.50%)" cards (08:48 and ~08:54 BKK) and asked "is this a new fee?"

**ANSWER: the rate is not new. 7.50% has been live since 7 Sep 2026. Neither contract in those cards is new either.**
- `0xbcb96b15dc2246d242c879316c86e25e846ad5fb` is the **7 Sep manager** already on file above (deployer nonce 71, 22,360 bytes).
- `0x83551a233f6d7b3cb4defbf55abfd887c6811ce4` is **deployer nonce 73, 29,283 bytes** (their largest manager ever), `feeBps()=750`, `treasury()=0x266311a44e70363b3d168bcb38cf38cbf2ccc51f`, `owner()=0xb1c2bbf86e557ecdc1812f75ae3fe973e5ec9e69`. Real and genuinely a new contract, but **first announced 13 Sep 15:35 UTC**, four days before these cards. Same rate, same treasury, same owner. Nonce 72 `0x56a6e3a7fe1ce9263be6441dd6e4a587683e75e7` is 4,950 bytes with no `feeBps`. Nonces 74 to 77 have no code yet.
- Alert-log receipts: 0x83551a announced 13 Sep 15:35 and 15:37, **again 16 Sep 23:48, again 17 Sep 01:48**; 0xbcb96b announced 13 Sep 16:16, **again 17 Sep 01:54** (all UTC).

**⚠⚠ ROOT CAUSE, AND IT IS THE ZERO-DEFAULT TRAP AGAIN, THE SAME ONE THAT COST AN ERROR ON CEREBRO THE SAME WEEK.** `fee_switch_snapshot()` built `feeBps_by_manager` with `if v is not None: fbs[m] = v`, so **a single failed `eth_call` dropped a manager out of the snapshot entirely**. The next cycle it came back, `m_ not in pf` was true, and the bot announced a four-day-old contract as new. The same function did `(as_int(b) or 0) / 1e18` on the burn-address balance, so **a failed read booked `delta_burned = 0`, and the next successful read looked like a 4,320,291 DELTA burn**. That fired **two fake "BUYBACK AND BURN" cards, #4 on 16 Sep and #5 on 17 Sep 01:16 UTC**, both quoting the cumulative lifetime total as a single burn. **A FAILED RPC READ IS NOT A ZERO AND IS NOT AN ABSENCE. Every bot that diffs two snapshots must raise and skip the cycle, never render a missing field as 0 or as "gone".**

**PATCH SHIPPED (deltaalert.py, backup `deltaalert.py.bak-20260917-021926`, syntax checked with `ast.parse`, service restarted 02:19:29 UTC, active, no alert since).** Four changes: (1) per-manager `feeBps` retries 5x and **raises** on failure, which the existing handler turns into "snapshot error, skip cycle"; (2) `balanceOf(dead)`, `balanceOf(treasury)` and the treasury ETH balance do the same instead of defaulting to 0; (3) **"NEW FEE CONTRACT SEEN" now keys off a persistent `state["fee_seen"]` set, not a one-cycle snapshot diff**, so an address can only ever be announced once; (4) `fee_seen` seeded with all 73 known addresses and the two bogus 4,320,290.72 rows deleted from `burn_log`, which is back to the three real burns (2,414,692 / 912,097 / 993,449).
- Method note: heredocs through `ssh` mangle quotes, so the patcher was written locally, **base64-encoded, piped through `ssh` and decoded on the droplet**. That is the reliable transport and should be the default for any droplet patch.

**⚠ THE REAL NEWS IN THE SAME DATA, AND IT IS NOT THE FEE: THE BURN PROGRAM HAS MISSED ITS SLOT.** Burns ran 7, 10, 13 Sep on an exact 3-day clock. **Today is the 17th and the burn address still holds exactly 4,320,291 DELTA, unchanged since 13 Sep. Four days, no burn.** Treasury holds **0.0104 ETH and 38,821 DELTA** (fee dust, under the 100,000 burn threshold), so it is refilling rather than sitting on an unspent buyback. Price $0.01568, cap $15,616,364, **+37.3% 24h**. **The falsifier named on 13 Sep was "it can stop on any cycle with no warning". One missed slot is not yet a stop, but the 20th is the date that settles it.**