Argus on Arc at day 7: 0 of 300 pools hold over 1,000 USDC

On September 16, 2026, the launch day of the Arc mainnet, the network's Uniswap v4 PoolManager registered 99,879 new pools —83,702 opened by the Argus Portal— and by day 7, none of the 300 pools in a random sample had more than 1,000 USDC inside —read on-chain at block 22,340,162, on September 23 at 11:27 UTC; the largest held 860.50— and only 5 (1.67%) had been traded in the 24 hours prior to 10:55 UTC. Argus is the launchpad (token factory with included pool) that dominated Arc's first day, the Circle network where gas is paid in USDC; a pool here is a trading pair, each with its token and USDC. Its value proposition is "locked liquidity": the 1 billion tokens from each launch go into a single Uniswap v4 position held by a contract with no withdrawal function. Argus locks the token, but does not guarantee the USDC: exit liquidity (the USDC a seller can take) only enters when someone buys, and the ~$2,481 figure that aggregators show for these pools is the entire supply valued at the opening price, with between 0.00 and 2.27 USDC on the other side. What follows is the on-chain census, the contract mechanics, the day 7 measurement with two definitions of liquidity, the contrast with pump.fun's 12.07% at seven days, and the precedent of Noxa and Pons on Robinhood Chain.

What does the Argus Locker on Arc actually lock?

Each Argus launch is a single transaction, and the repository published on September 16, 2026 (declared MIT in the README, and which the repository itself describes as a simplified reference for the deployed contracts; of these, Portal v7 is verified on arc.etherscan.io as ArgusV4Portal7) allows it to be followed step by step: the token is cloned with 1 billion units, a revenue splitter (the contract that distributes fees between the creator and Argus) and a hook specific to that token are deployed (the hook is the contract that Uniswap v4 executes before or after each swap in the pool), the pool is initialized at the opening price set by the creator, the entire supply is deposited into a single position between two ticks (the price steps between which a position offers liquidity), the Locker that holds that position is deployed, and the CurveOpened event is emitted. Fixed pool fee at 1%. The README summarizes it: "there is no virtual bonding curve: the entire token supply is deposited into a single v4 liquidity position, above the opening price."

A position placed above the opening price contains only tokens: liquidity.base is the inventory, liquidity.quote is the USDC that a seller can take, and the "total" value on the dashboard is the inventory multiplied by the opening price. The lock guarantees that no one withdraws either the tokens or the USDC; it does not guarantee that there is any USDC.

The Locker in the repository has five immutable variables and a single public function, harvestFees(), which anyone can call and which sends the position's fees to the splitter. The code comment: "Deliberately NO OTHER function. No withdraw, no rescue, no way to modify the position, no admin function of any kind." According to that code, an LP rug —withdrawing liquidity and leaving the pool empty— is not possible for either the creator or Argus; the bytecode of the deployed Lockers is not verified in the explorer and has not been read for this piece.

How many pools did Arc open on Sep-16-2026 and how many belong to Argus?

The census was conducted using two different RPCs over blocks 21,068,653 to 21,239,146 (the 170,494 blocks of September 16 in UTC), filtering for the Initialize event of PoolManager 0x8366…0951 and the CurveOpened event of Argus Portal v7 0xB021…97Da. Both counts matched.

Each Argus launch deploys its own hook: of the 84,244 hook addresses, 83,702 are one hook per Argus token. The Dune dashboard "the arc trenches" (aggregator, daily refresh; reading from September 24, 2026, at 20:01 UTC) counts 83,751 Arguspad launches on September 16 —49 more than the CurveOpened events of Portal v7, a 0.06% difference— and 97,025 "tokens" across its 19 launchpad rows (the total varies between refreshes: 95,897 on the 23rd) compared to the 99,879 pools in the census: the difference lies in pools without hooks and those from other protocols on v4. The same dashboard shows the decline in the following days: 48,559 Arguspad launches on the 17th and 14,172 on the 18th, an 83% drop in two days, with a declared volume of $202.35 million on the 16th, 42.28 on the 17th, and 23.22 on the 18th. Day one volume was discussed in Arc vs Tempo. According to DefiLlama (reading from September 24, 2026, at 19:58 UTC, identical for those days to the reading from the 23rd at 11:10 UTC; the aggregator overwrites past days), the network-wide DEX volume fell from $125.59 million on the 17th to between 40.46 and 50.45 daily from the 19th to the 22nd; the figure for the 16th (104.10) does not include Uniswap v4 and is not comparable to Arguspad's $202.35 million.

Why does DexScreener show $2,481 of liquidity in Argus pools with less than 1 USDC?

The ~$2,481 figure repeats pool after pool because it is the same calculation: 1,000,000,000 tokens multiplied by the opening price of $0.00000248 ($2.48 per million tokens). Since the entire supply is in the pool, "liquidity" and fully diluted valuation are the same number. DexScreener values both sides of the position, which is its definition. In the reading from September 19 at 10:55 UTC, 18 of the 28 pools in the sample indexed by DexScreener showed liquidity.usd between $2,481 and $2,485 with liquidity.base of 999 to 1,000 million tokens and liquidity.quote between 0.00 and 2.27 USDC. On September 23 at 10:55 UTC, the aggregator only returned 5 pools from the sample, and the only one with an intact supply remained at $2,481.94 with 0.60 USDC inside.

These are the five pools the aggregator returned in the two passes on September 23 (10:55 and 11:06 UTC), which were identical; on September 19 at 10:55 UTC, it returned 28 from the same sample. Only WORKCHAIN and DUKE are from Argus. WORKCHAIN (999.76 million tokens, 0.60 USDC, and one swap in three days) is the typical case: its "liquidity" of $2,481.94 is the opening price. DUKE is the only one with more than 400 USDC and 849 million tokens still in the pool. Across the full sample, the naive threshold —liquidity.usd over $1,000 and some volume— was met by 3 out of 300 pools (1.00%: WORKCHAIN, ARCADE, and DUKE); the exit threshold —more than 1,000 USDC inside— was met by 0 out of 300 in the on-chain reading at block 22,340,162 (details in the next section). On September 19, the gap between the two definitions was wider: 23 pools (7.67%) passed the naive threshold and 0 passed the exit threshold—23 "survivors" that existed only in the inventory valuation.

How many pools from the Arc cohort are traded at day 7 and how much USDC can they return?

The sample consists of 300 pool identifiers extracted from the census using seed 20260919, the same for both day 3 and day 7; 245 of the 300 (81.7%) are Argus launches. For each pool, DexScreener was consulted and Swap events from the PoolManager were counted in the 172,800 blocks prior to the measurement (24 hours at 0.5 seconds per block; 24.4 hours at the measured 0.508). Definitions set before measuring: traded means having at least one swap in those 24 hours; exit liquidity is liquidity.quote; alive means meeting both with more than 1,000 USDC.

The USDC side was also read on-chain for all 300 pools, independent of the aggregator's index, at block 22,340,162 (September 23, 11:27:50 UTC). The v4 PoolManager is a single contract with a single USDC balance for all pools —3,704,273 USDC that day— so the USDC "of a pool" is the amount implicit in its liquidity distribution at the current price, the same definition DexScreener uses for liquidity.quote: ticks for each pool were taken from their ModifyLiquidity events (1,127 in the sample), the actual state of those ticks and the price were read with extsload, and integrated with concentrated liquidity math. In all 300 pools, the sum of liquidity per tick matched the contract's active liquidity, and in the 5 that DexScreener did index, the result matches the aggregator to the fourth decimal place (DUKE: 439.6885 vs 439.6884). Result: 295 pools have a USDC side (282 with the ERC-20 and 13 with Arc's native USDC, the same asset with 18 decimals) and 5 are pairs between two tokens without USDC, 3 of them launched by Argus with another quote currency. None exceed 1,000 USDC; the largest, UAG, an Argus pool not indexed by the aggregator, holds 860.50; 4 exceed 100; 274 have 1 USDC or less, and 198 have exactly 0. The 295 total 1,810.09 USDC, with an average of 6.14 and a median of 0; the 242 Argus pools with a USDC side total 1,410.02, and 233 of them have 1 USDC or less.

On day 7, the 5 pools with swaps in 24 hours were exactly the 5 indexed by DexScreener in both passes, and the 33 swaps of the day were distributed as 20, 8, 3, 1, and 1. Over 72 hours, 9 pools traded with 203 swaps, 140 of them in LIFT and MDOG, and 4 of those 9 with a single swap. For the 245 Argus pools in the sample, the figure is lower: 2 assets in 24 hours (0.82%) and 4 in 72 hours (1.63%). The 5 active assets on day 7 are among the 99 that traded on their creation day; of the 9 active assets in 72 hours, 8 were as well, and 1 had its first swap later. On September 19, there were 28, almost all with a single swap: between day 3 and day 7, activity fell to less than one-fifth.

The thesis was formulated with two falsification tests before measuring.

1. (A) If 3 or more of the 300 pools (≥ 1%) had more than 1,000 USDC of exit liquidity on September 23, the liquidity part was false: the result is 0 out of 300, read on-chain at block 22,340,162 (11:27 UTC), with a maximum of 860.50 USDC.
2. (B) If 37 or more (≥ 12.07%) had at least one swap in the 24 hours prior to that measurement, the comparison with pump.fun was false: the result is 5 out of 300 (1.67%), with a 95% confidence interval of 0.71 to 3.84%; with the 72-hour window, 9 out of 300 (3.00%, interval 1.59-5.60%). Both parts hold.

What limits the Arc cohort measurement and how is it repeated?

The DexScreener index returned 28, 10, and 5 pools from the same sample in three queries (September 19 at 10:55 and 14:32 UTC; September 23 at 10:55 and 11:06 UTC, these two identical); therefore, the USDC side for day 7 was read on-chain for all 300, and the aggregator remains only as a contrast; on day 3, the USDC side of non-indexed pools was not measured, and 73 of those 272 had traded on September 16, so not being in the index does not equate to being inactive. No pool with a swap was left out of the index in any reading. The on-chain reading excludes unclaimed liquidity fees, just like the aggregator. With 0 out of 300, the upper limit of the 95% interval for part A is 1.26%: the test was set on the observed count (3 or more) and is reported as such, without claiming the true proportion is zero. Repeating the measurement requires the same seed (20260919) on the census of 99,879 identifiers, the same block windows for swaps, and the same fixed block for the USDC side.

What does it cost to buy and sell an Argus token?

The LaunchHook for each token carries two immutable taxes, buyTaxBps and sellTaxBps. The Portal and the hook in the repository (published September 16, 2026) only check that neither exceeds 1,000 basis points (10%) and that both are not zero: the contract interval is 0 to 10% per side with at least one side non-zero, and the "1% to 10%" in the documentation is the range offered by the creation interface. The tax is deducted in USDC on each swap (USDC is the quote currency in 94.7% of the census pools) and is added to the pool fee, the 1% of each swap charged by the Locker position. A round trip pays between 2.01% (contract: one side at 0.01%, plus 1% pool fee each way), 3% with the interface minimum, and 22% with both sides at 10%. The splitter distributes 10% to Argus and 90% to the creator: for every 1,000 USDC of volume with a 5% tax per side, 50 USDC go to the splitter, 45 to the creator, and 5 to Argus, plus the 10 USDC pool fee which follows the same path.

The opening surcharge, or snipe tax, is calculated with a bit shift: 9900 >> ((elapsed * 14) / 3), with elapsed in seconds since launch. At second 0, the surcharge is 9,900 basis points (99%, the combined cap); at second 1, it drops to 618 points (6.18%); at second 2, it is 19 (0.19%); at second 3, it is zero. With 0.508-second blocks, the first block after launch and, if it falls within the same second, the second block pay 99%; the Portal and the splitter are exempt, covering the creator's own purchase in the launch transaction. The surcharge only affects the first block; creator revenue comes from subsequent volume. WORKCHAIN, with 999.76 million tokens intact and 0.60 USDC on September 23 at 10:55 UTC, produces nothing; LIFT, with 20 swaps that day and a $219.76 valuation, does.

Is the Arc cohort worse than pump.fun at seven days?

CoinGecko Research published on June 23, 2026, the lifespan distribution of 18,675,645 pump.fun tokens created between January 14, 2024, and June 18, 2026, excluding those that never had a trade and defining lifespan as the days between creation and the last trade on the pump.fun bonding curve (a contract that sells the token at an increasing price until the curve is exhausted and liquidity migrates to a DEX).

The 12.07% is the sum of the rows after day 7 in the CoinGecko table; 80.37% of tokens had their last trade within 48 hours, 7.55% survive more than 30 days, and 4.55% more than 90. CoinGecko itself warns that its metric does not capture trades after graduation to a DEX, so 12.07% is a floor. Its metric counts any token that traded again after day 7; this cohort's metric is a 24-hour window on day 7, which is stricter, and that is why a 72-hour window was added: 9 out of 300 (3.00%) on September 23, also below. The base also differs: CoinGecko excludes tokens without any trades, while in the Arc sample, 201 out of 300 pools (67.0%) did not have a single swap on their creation day. Recalculated for the 99 that traded on their creation day, the 5 with activity at day 7 are 5.1% (the 8 out of 72 hours that also traded that day are 8.1%), still below 12.07%.

There is also a design asymmetry. On pump.fun, a token only gets a pool with real liquidity if the curve is exhausted, and most die before then, as detailed in the analysis of the $1.08 billion in revenue generated by pump.fun. On Argus, all tokens are born with a pool and a locked position; the USDC arrives, if it arrives, with the first purchase.

What happened on Robinhood Chain with Noxa and Pons before Arc?

Robinhood Chain reached public mainnet on July 1, 2026 —what is real and what is marketing in Robinhood's L2 counted 13,900 contracts in the first few days— and its dominant launchpad was Noxa: about 60,000 tokens launched until July 11 according to Tom Wan's count on July 12 and nearly $12 million in accumulated fees according to CoinDesk. On July 11, Noxa paused launches citing bots that copied tokens every hour; on the 13th, its website stopped responding, and on the 14th, it announced it was giving 100% of revenue to creators; CASHCAT, the largest token born there, fell more than 33% in 24 hours. Pons filled the gap starting July 13 and on September 3 generated $5.95 million in revenue in one day according to Decrypt.

The precedent identifies the single point of failure. On Robinhood Chain, the leading launchpad shut down on the 11th and activity moved to another within 48 hours; Noxa tokens simultaneously lost their factory, their website, and their price reference. On Arc, Argus opened 83.8% of day-one pools, and the other 18 launchpads on the Dune dashboard share the remaining 13,274 (sum of rows from September 24 at 20:01 UTC). A network whose TVL (the value deposited in its protocols) is provided by two lending markets (Morpho and Aave V4, where three addresses provided 98.1% of the USDC, 90.4% of the total on September 17 according to the DefiLlama snapshot cited in Arc vs Tempo) and whose transactions are provided by a launchpad has two economies that do not touch. Which exchange L2 lasts with that structure was compared in Robinhood Chain, Base, and Ink.

What does the 0x report say about v4 hooks and what changes in Argus hooks?

On September 14, 2026, 0x published an analysis of 84,163 Uniswap v4 hooks across six chains, with data as of September 11: 19.4% safe; 54.2% malicious; 26.4% probably malicious, where malicious means announcing a different quote than what the user receives. Arc is excluded from the study due to timing —the mainnet opened five days after the cutoff— and in a single day, it added 84,244 hook addresses, more than the entire 0x sample. Argus hooks do not fit the 0x definition: the tax is public and immutable, and the contract exposes totalFeeBps(bool isBuy). The risk of incorrect quoting exists if a router assumes only the 1% pool fee: it will announce a price up to 10 points better than reality in each direction. Why Uniswap is pushing v4 is in the analysis of fee activation and UNI burning.

What can a reader check in an Argus pool before trading?

In the DexScreener pair sheet, liquidity.quote tells how many USDC are inside the pool and liquidity.usd tells how the inventory is valued: cents in the first and $2,481 in the second mean no one has bought yet or everything bought has already exited. In the hook, buyTaxBps and sellTaxBps are public readings that the reference code declares immutable. What no field provides is how much of a pool can be sold without crashing the price: with 440 USDC inside and 151 million tokens in circulation (849 million remain in the pool), as in the DUKE pool in the table on September 23, the answer is approximately 440 USDC.

The cohort will be measured again at 30 days, on October 16, 2026, with the same sample and definitions. On September 23, at block 22,340,162, USDC did not exceed 1,000 in any of the 300 pools in the sample, did not reach 100 in 296, and was 1 or less in 274; in the 245 from Argus, in 233.

Sources and links: Argus — argus-world repository (README and Portal, LaunchHook, Locker, RevenueSplitter contracts) · Argus — Portal.sol (tax validation and createLaunch) · Argus — LaunchHook.sol (immutable taxes and snipe tax formula) · Argus — Locker.sol (no withdrawal function) · arc.etherscan.io — Portal v7 verified as ArgusV4Portal7 · Arc Docs — RPC and chain ID 5042 · Arc Docs — contract addresses (18-decimal native USDC and 6-decimal ERC-20) · DexScreener — pair API reference · CoinGecko Research — average lifespan of pump.fun tokens (Jun-23-2026) · Dune — CoinGecko query 7755022 · Dune — "the arc trenches" dashboard (launches by launchpad) · DefiLlama — Arc DEX volume · 0x — "Uniswap v4 hooks were a mistake" (Sep-14-2026) · CoinDesk — Noxa gives away 100% of its revenue (Jul-15-2026) · Tom Wan — 60,000 tokens launched from Noxa (Jul-12-2026) · Decrypt (via Yahoo Finance) — Pons and its $5.95 million daily (Sep-3-2026) · Datawallet — Pons mechanics (curve, graduation, fees)