Pre-market withdraw
Exit works in the first session of the day, with session fees and flow caps in force. TSLA / USDG pool, through the Uniswap v4 PositionManager.
0x3a6144b81e988cdda55994fb6eafdc55067fb26c1694c50765fcf271d5628bfc
Open on BlockscoutExit liveness proof
No BELL policy can hold back a withdraw. You do not have to take our word for it: this page shows how to check it from a hook's address, runs that check on every pool, and links a real withdraw we made on mainnet.
Chain 4663 / hook mask 0x2cc0
0xf9692a364e0edf76bba12e54c3eba599f436acc0
The lowest 14 bits of the address are the hook's permissions. The two bits before them are part of the address and mean nothing. A filled bar means the hook is called at that step. The three marked exit are the ones that would touch a withdraw, and they are empty.
In Uniswap v4 a hook's permissions are not stored anywhere you have to trust. They are the low 14 bits of the hook's own address. The PoolManager calls a hook callback only when its bit is set, so a hook whose remove-liquidity bits are clear is never called while liquidity is removed. That can be read from the address alone, without reading any source code.
BELL hook addresses end in cc0. The digit before it varies, because the two highest of those sixteen bits are not permissions. The fourteen that are read 0x2cc0, which sets five permissions and clears the three that touch removal.
| Bit | Permission | In 0x2cc0 |
|---|---|---|
| 13 | beforeInitialize | set |
| 12 | afterInitialize | clear |
| 11 | beforeAddLiquidity | set |
| 10 | afterAddLiquidity | set |
| 9 | beforeRemoveLiquidityexit | clear |
| 8 | afterRemoveLiquidityexit | clear |
| 7 | beforeSwap | set |
| 6 | afterSwap | set |
| 5 | beforeDonate | clear |
| 4 | afterDonate | clear |
| 3 | beforeSwapReturnDelta | clear |
| 2 | afterSwapReturnDelta | clear |
| 1 | afterAddLiquidityReturnDelta | clear |
| 0 | afterRemoveLiquidityReturnDeltaexit | clear |
The three rows marked exit are bits 9, 8 and 0. A hook with any of them set could act when you withdraw. None of BELL's does.
The list below comes from the BELL keeper's /pools, and the arithmetic runs in your browser. If the keeper cannot be reached, it shows the list committed to the repository instead and says so.
Showing the list committed to the repo, generated 2026-11-06. Reading the live list…
| Pool | Hook | Flags | Exit bits | Result |
|---|---|---|---|---|
| AAPL / USDG | 0xf9692a…36acc0 | 0x2cc0 | none set | PASS |
| AAPL / USDG | 0x6c31e4…e2acc0 | 0x2cc0 | none set | PASS |
| AAPL / USDG | 0xc1ada0…886cc0 | 0x2cc0 | none set | PASS |
| NVDA / USDG | 0x89d479…27ecc0 | 0x2cc0 | none set | PASS |
| TSLA / USDG | 0x6657dc…0decc0 | 0x2cc0 | none set | PASS |
| TSLA / USDG | 0xeece9e…d06cc0 | 0x2cc0 | none set | PASS |
6 hooks, 6 PASS, 0 FAIL. 6 carry the BELL mask 0x2cc0.
The deployed bytecode also reports its own permissions through getHookPermissions(). Running the script with --rpc confirms that each hook has code and that what it reports matches what its address says.
Do not trust the table either. Paste a hook address and see its fourteen bits.
Paste any Uniswap v4 hook address. Nothing is sent anywhere.
One real withdraw from a BELL pool on Robinhood Chain mainnet, through the Uniswap v4 PositionManager. Open the transaction and check it yourself.
Exit works in the first session of the day, with session fees and flow caps in force. TSLA / USDG pool, through the Uniswap v4 PositionManager.
0x3a6144b81e988cdda55994fb6eafdc55067fb26c1694c50765fcf271d5628bfc
Open on BlockscoutThe contract tests below run in the Foundry suite. The committed hook list is checked on every push, and the live pool list every six hours. Either check fails the build if any hook has an exit bit set.
testHookPermissionsAreExactAndReturnDeltaFlagsAreFalseThe hook reports exactly mask 0x2cc0, no remove-liquidity flag, and its address carries those bits.testPublicInKindExitWhilePoolAndAssetPausedPool and asset both paused through the safety controller: the owner still exits.testPublicInKindExitWithStaleHealthAndCalendarHealth and calendar data gone stale: the owner still exits.testPublicInKindExitWithAllBellDependenciesUnavailableEvery BELL dependency reverts on read: the owner still exits.testPublicInKindExitClosedLatchedAndExhaustedBudgetsMarket closed, closure latched and every flow budget spent: the owner still exits.testFuzzOwnerCanAlwaysExitAfterAnySequenceOfBellEventsThe fuzzer combines pauses, dead dependencies, stale time, a latched closure and spent budgets in any order. After every step the owner removes all liquidity in kind and BELL's state is unchanged.testOwnerExitsAfterEveryEventAtOnceEvery kind of event stacked on one pool, in a fixed order, so a regression names itself without a seed.The randomised test draws up to 24 events per run and 1,500 runs per pass. It checks the exit after each event on a snapshot it then rolls back. It does not cover swaps, which BELL is allowed to stop, or the Uniswap contracts themselves.
Node 24 runs the script directly, with no install for the basic check.
git clone https://github.com/belldex/bell cd bell # every pool the keeper lists node scripts/verify-exit-liveness.ts # one or more addresses you choose node scripts/verify-exit-liveness.ts --hook 0x6657dc160e05c0ff6afe2122ca7d8244390decc0 # also check code exists and getHookPermissions() agrees (needs an RPC URL and viem) node scripts/verify-exit-liveness.ts --rpc https://your-robinhood-chain-rpc
It prints one row per hook with PASS or FAIL and exits with 1 if any hook can touch a withdraw. The source is scripts/verify-exit-liveness.ts and the arithmetic is a few lines in hook-flags.ts.
It means no BELL policy can hold a withdraw back: no pause, no closed market, no stale feed, no spent budget and no admin action in the BELL contracts stops the owner of a position from removing liquidity.
It does not mean more than that:
Exit liveness is about getting your tokens out. This section is about what an LP keeps while the market is closed. We replayed 20 real weekends of TSLA, NVDA and AAPL, 2026-06-05 to 2026-10-30, against two pools that start with the same liquidity, the same range and the same price: a standard Uniswap v4 pool with a fixed 0.30% fee, and a BELL pool with the Opening Bell Auction.
This replay is a simulation, not a live weekend result. A live side-by-side test of a standard pool and a BELL pool over a real weekend has not been run. What has run on Robinhood Chain mainnet is listed under the replay.
| Over 60 stock-weekends | Standard pool | BELL pool |
|---|---|---|
| Average LP result against holding | -0.42 bp | -0.25 bp |
| Largest loss in one stock-weekend | -3.48 bp | -1.86 bp |
| Gap of 1% or more (16 rows), average | -1.41 bp | -0.78 bp |
| Gap under 1% (44 rows), average | -0.06 bp | -0.06 bp |
One basis point (bp) is 0.01% of the starting liquidity, which was 100,000 USD in each pool. BELL did better in 16 of 60 rows, the standard pool did better in 0, and 44 were a tie. The ties are the weekends with a gap under about 1%: there both pools end at the same price with the same fees. The difference appears when the gap is large, which is when the closed-session fee and the flow caps act.
In the 16 stock-weekends with a gap of 1% or more, the LP in the standard pool lost -1.41 bp on average and the LP in the BELL pool lost -0.78 bp, which is 45% less. The LP still pays for most of the move in both pools, because the price has to reach its new level either way. The difference comes from fees: arbitrage while the market is closed is charged 1.00% in a BELL pool and 0.30% in the standard pool, and the fee goes to the LP. On 100,000 USD of liquidity, that is a few dollars per stock-weekend.


What the simulation assumes, and where it stops:
What has run on Robinhood Chain mainnet so far, each one a transaction you can open:
TradeCapExceeded: transaction.The one real weekend on record is 7 and 8 November. Three funded BELL pools held liquidity through it, and none of them took a swap, so no LP in them paid anything for the weekend. Prices below were read on 9 November at 02:57 UTC; the Friday close is the last post-market bar on 6 November.
| Pool | Pool price | Friday close | Difference | Swaps since creation |
|---|---|---|---|---|
| AAPL/USDG | 325.21 | 342.65 | -5.1% | 0 |
| NVDA/USDG | 207.05 | 188.70 | +9.7% | 0 |
| TSLA/USDG | 350.34 | 359.40 | -2.5% | 0 |
Read this carefully. The differences come from the price each pool was seeded at, not from the weekend, and nobody has traded to correct them. So these pools show that a BELL pool takes no flow while it sits unpriced, and they do not show that a BELL LP loses less than an LP in a standard pool: there is no standard pool beside them and there was no arbitrage to lose to. What has not run: a live weekend measured side by side with a standard pool, a mainnet withdraw while the market was closed, and an Opening Bell Auction on mainnet with orders in it (the cycle on 9 November closed with none). The amounts are small, so they show that the rules work, not what an LP keeps.
The script, the prices, the result of every row and the charts are in the public repository:research/replay. To rerun it, run forge test --match-contract WeekendReplay -vv from the contracts folder.