Breach Zero rehearses a launch every six minutes

We built the first pieces of BREACH, the launchpad we plan to run inside the city. There are two: a rules engine that decides who gets what in a launch, and the Breach Zero rehearsal board, which runs those rules in public on made-up numbers.

Breach Zero is a rehearsal with no token and nothing to buy. The on-chain program a real launch would need is not built, and nothing has been audited.

The rules engine

The engine is pure functions, and every amount in it is an integer.

The core is the batch. People commit during a window, and when it closes, everyone who gets a fill pays the same price. Each commit is first held to its passport's cap. If the batch is oversubscribed, fills are cut pro-rata and the rest is refunded. Nobody is picked by lottery. The launcher commits in the same batch and is capped at 3% of its tokens. With one commit per passport, arrival order changes nothing. A property test shuffles 400 random batches and checks that every fill comes out the same.

After the batch, 25% of each fill unlocks at T0 and the rest unlocks evenly over 45 minutes. When trading opens, an anti-snipe fee starts at 50% and falls to 1% over 120 seconds, rounded up so it never undercharges.

The Furnace is the share of fees set aside to be burned. Off the top, before anything else is paid, the engine gives it:

• 0.15% of each curve trade
• a third of the pad's share of LP fees
• all slot fees
• anti-snipe fees above the 1% floor

When a curve trade fee, or the pad's share of an LP fee, does not divide evenly, the leftover goes to the Furnace. In a later test of the trading setup we plan to use first, the Furnace's cut of curve trades came out at 0.09%, so the 0.15% rate is not settled.

The Verdict grades a launch's process from A+ to F. The score is out of 100: 40 points for how many people committed, 30 for how small the largest fill is, and 30 that fall to zero as the estimated share held by linked wallets reaches 20%, scaled by how confident that estimate is. Four hard checks sit on top: no team allocation, a launcher at 3% or less, verified locks, and an LP that stays locked. Fail any one and the grade is capped at D. The grade never changes fees.

The engine has 103 tests. A red team checked it against our written spec and found five bugs, all fixed with regression tests. Two of them:

• When the launcher went over its cap in an oversubscribed batch, the engine cut the launcher but did not give the freed room to the other committers.
• A batch set up with zero supply would have accepted commits and handed out no tokens. It now refuses to run.

The rehearsal board

The board runs the same engine on a shared six-minute loop aligned to UTC: a 60-second omen that announces the launch, a 90-second commit window, 6 seconds of clearing and 180 seconds on the curve, then a 24-second countdown to the next one. Committers are generated from the cycle number, so every visitor sees the same rehearsal at the same moment. Amounts are in SIM, a made-up unit with no market or contract behind it. The tape reads "REHEARSAL · NO TOKEN · NOTHING TO BUY".

You can enter your own commit during the window and see your fill after clearing. The board sends no data, so your commit stays on the page. The Verdict grades each rehearsal once it clears. Rehearsals alternate: one graduates and the next is marked dead, so the board shows both endings. The board shipped with 22 tests of its own.

Still owed

Our first look at the on-chain side found the batch needs an escrow program of our own, which will need an audit. Passports are not built either. The final Verdict formula and an appeals process will be published before the first real launch. Any number here may change before then. The token has not launched. The disclaimer is on the token page.