commits · the thesis

It is not writing code that is slow. Writing code is fun, doable, well-paid, frequently delegated to a fast LLM. What is slow is getting code merged. Maintainers triage backlogs in the thousands. Drive-by reviewers leave 👀 emojis. Drift kills more PRs than bugs do.

The economics are upside-down: writing earns reputation, reviewing earns nothing. The labor market for review doesn't exist because the only currency is favor, and favor doesn't scale.

The trick is not paying reviewers to approve PRs. That gives you rubber stamps. The trick is paying them to commit their judgement onchain with stake — and slashing them when they're wrong. The mechanism rewards correct calls, not friendly ones.

This is the same insight that makes prediction markets work and chain consensus work. Skin in the game beats vibes every time.

Repo owners deposit $MERGE into a pool against an open PR. Reviewers stake $MERGE to vouch. On clean settle, the pool releases proportional to stake. On revert within 256 blocks, vouchers are slashed and the slash redistributes to anyone who explicitly rejected the PR.

Two-sided economy. Owners pay for attention. Reviewers pay for the right to earn.

Base because the gas envelope makes per-PR review economically sensible. Every step of the loop fits comfortably under $0.05.

Gitlawb-native because the agent web is going to ship code at volumes humans can't review. Stake is the only filter that scales.

Agent-friendly because reviewers are wallets. Any agent that passes the gitlawb verifier can vouch on the same terms as humans.

• supply: 1,000,000,000 $MERGE at genesis · no emissions · no inflation
• burn: 2.5% of every settled pool burns permanently · treasury keeps 2.5%
• distribution: 50% liquidity / 30% community + reviewer incentives / 15% protocol treasury / 5% team (4-yr vest)
• governance: none. parameters are fixed at deploy. fork the contract if you don't like it.

The 2.5% treasury cut on every settle compounds into a protocol-owned pool. That pool can be voted (by token-weighted snapshot) to top up review pools for high-priority OSS repos — the leftpads of the world, but before the next leftpad happens.

This makes $MERGE useful even for repos that can't afford to fund their own pools. Critical infra gets reviewed. Maintainers don't pay out of pocket. The cost is socialized across everyone using the protocol.

• p0 — token deploy · $MERGE on Base · CA published · LP locked
• p1 — protocol mainnet · vouch + slash + settle contracts · webhook bridge for github / gitlab
• p2 — gitlawb-native · zero-config opt-in for any gitlawb repo · agent verifier hookup
• p3 — composability · treasury subsidy program · reviewer dividend split · partial-stake voucher pools

• oracle risk · CI is the oracle. compromised CI = bad settlement. mitigation: settlement window + revert detection
• sybil reviewers · partial mitigation via reputation gating on high-pool PRs · open problem at the long tail
• collusion · owner + reviewers coordinate to drain pools. mitigation: random subset of vouchers slashed extra if everything settles too clean too fast (heuristic, work in progress)
• maintainer bypass · maintainers might just merge without checking pool. mitigation: pool only releases on owner-confirmed merge tx, not on any merge

Every merge is an act of judgement made by a few humans (or agents) on behalf of everyone who will run the code. $MERGE is the first attempt to give that judgement an explicit price, an explicit incentive, and an explicit cost when wrong.

The merge happens when we agree. We agree when the stakes are on the table.