The Bundle That Cannot Sell: How Mosh Turned Bundle Cat's Launch Wallets Into Its Market Maker

Bundle Cat, the first token launched through Mosh, a launch and market-making system built by UV Labs on the Pons launchpad on Robinhood Chain, has been trading for a little over a week. Mosh has now published its first account of how the launch went and how the machinery underneath it works. The short version: the wallets that bought at launch cannot sell, and a set of AI agents trades the tokens they hold instead.

According to Mosh, the raise that funded the launch was filled with 8 ETH, of which 4 ETH bought the opening bundle. The launch earned that 8 ETH back in fees within the first minute, and about 72 percent of the supply was locked in the agent vaults. Since then the agents have bought 1.22 million BUN for 2.4 ETH and sold 1.26 million BUN for 4.2 ETH. CoinGecko lists the most active pair as BUN against WETH on Uniswap v4 on Robinhood Chain.

The token by the numbers

BUN went live on Pons at 15:37 UTC on 2 September, according to GeckoTerminal, which puts the token thirteen days old. Its first hour closed at about $0.0033. The price reached $0.017 the next day, fell to $0.011 on 4 September, and then climbed for a week to an all-time high of $0.0526 on 10 September, CoinGecko records. It has given back more than half of that since and traded at about $0.022 on the morning of 15 September, up 25 percent over seven days.

At that price the market cap is $6.3 million on a circulating supply of 286 million BUN, and the fully diluted value is $22 million on the total supply of 1 billion. The gap between the two is the vaults: 71 percent of the supply is not in circulation, which matches the 72 percent Mosh says it locked at launch. Volume across all pools was $1.7 million in the last 24 hours, with the deepest pool, BUN against WETH on Pons, holding about $450,000 of reserves and doing $546,000 of that volume.

GeckoTerminal counts 2,134 holders and puts 83 percent of the supply in the top ten addresses. Bebis's claim and that figure can both be true, because the concentration is in contracts, not people. The vaults hold most of the top ten, and the remaining 17 percent is spread across two thousand wallets with no single person above half a percent. Twenty-two pools list the token, most of them empty, which is normal for a Robinhood Chain launch: anyone can open a pool, and a handful of them carry all the trade.

The problem a bundle usually causes

On a launchpad, a bundle is a set of wallets that buys a large share of the supply in the launch transaction, ahead of everyone else. Whoever controls it is the largest holder from the first block and, sooner or later, the largest seller. Every trader in the pair is trading underneath that overhang, and the price is capped in practice by the moment the bundler decides to leave. On Pons, where the Ledger reported in August that most tokens stop trading within a day, that moment usually arrives early.

Mosh's answer is to keep the bundle and remove the exit. The tokens still get bought in size at launch, and the size still matters, because it is what gives the market a balance sheet. But the inventory goes into vaults that can be traded by software under limits and cannot be withdrawn by anyone.

What Mosh actually does

Mosh describes itself as three things: a capital formation mechanism, a token launch management system and a token lifecycle management system. Mechanically, it runs a small raise from funders to pay for the opening bundle. A whitelist is optional and lets the deployer decide who may contribute and how much. That capital then executes a sequence of actions in one go that cannot be front-run: buy on the curve, graduate it, add liquidity. The bundle is locked in the vaults in the same step.

Creator fees are paid back to the funders pro rata. That is the part that changes the incentive. A funder's return is the fee stream, not the sale of tokens, so the largest early holder has no reason to dump and no ability to.

The agents

The agents were built by UV Labs, which posts as @uv, on the firm's own AI harness. Mosh says they identify market regimes, reason about order flow, take in arbitrary event streams and run strategies through a custom high-frequency trading toolchain. For Bundle Cat, the job is market making from inventory. The agents place bids and asks off-chain that fire when a Uniswap tick is crossed, setting spread, size and skew from volume, volatility, inventory and distance from the mid price, among other inputs.

The strategy was first developed in simulation and is now being replaced, step by step, by one tuned on production data from the BUN market. Mosh intends to use the traces from future launches to train custom models for memecoin markets. The language model sits on top to reduce maintenance, and it is only switched on for markets with real volume, which is how the firm expects to run thousands of launches a day cheaply.

On the team's own figures, the agents have so far sold about 40,000 more BUN than they bought and taken in a net 1.8 ETH. Mosh describes their mandate as accumulating BUN while quoting both sides of the market. The first week's numbers show more of the quoting than the accumulating.

What the agents are not allowed to do

Mosh is direct about the reputation of market making in crypto and says its agents do not paint charts, do not wash trade and do not guarantee outcomes. For BUN, they trade back and forth to pull tokens off the market and supply liquidity during sharp moves. Every function is limited by a cap on market actions per period, with both an absolute ceiling and a ceiling set as a percentage of inventory. The agents cannot withdraw.

Since the launch, Mosh has added the ability to manage range liquidity, both inside the launch transaction and during normal operation. Planned agent functions include over-the-counter purchases, community NFT mints to build inventory, and access to perpetual futures and other instruments from inside the vault, under the same limits.

Why distribution is the point

Justin Bebis, UV Labs' chief executive, made the case on X on 8 September: "Launchpads took off because they allowed people to create markets in seconds. Mosh will take off because it allows people to create healthy markets in seconds. Look at $BUN distribution. It's a Pons token and only one person owns more than 0.5% of the supply."

That is the structural argument for a token launched this way. A supply with no large private holder cannot be sold down by one decision. The largest holder is the vault, and the vault is the market maker, so the biggest position in the book is also the one that never leaves. Mosh says the next step is to use creator fees to fund a second kind of agent that re-enters its markets as a cash-rich buyer, starting with BUN. The team calls these agentic liquid funds and describes them as buyback tokenomics with a trader attached, with the aim of making memes easier to underwrite and, eventually, of getting institutional capital to treat bundles as yield products.

What the mechanism does for a token is narrow and real. It removes two of the usual causes of death on a launchpad, the bundler's exit and the empty book, and it turns the launch inventory into standing liquidity. It does not create demand, and the agents are not designed to. Whether Bundle Cat is still trading at 14 days will say more than the first week has.

Pons first

Mosh says it is built to sit on top of any launchpad but is concentrating on Pons, the bonding-curve launchpad that has become the largest meme coin factory on Robinhood Chain. Bebis wrote that Pons needs Mosh more than any other launchpad on the chain, and that the aim is for Mosh to become a standard traders look for in a new pair, with launch discipline enforced at the smart-contract level rather than by trust. The system is meant to cover the whole life of a token, from raise and launch through sustainment and wind-down.

Discussion is in the project's Telegram group, The Mosh Pit, and the Bundle Cat token page is on mosh.trade, with the contract on Robinhood Chain's Blockscout.