Two weeks ago, Phantom crossed $10 million in revenue from a feature they didn't build. They just plugged into Hyperliquid's perpetual trading API, attached their "builder code" to every order, and collected 0.05% on $20 billion in trading volume.
The interesting part wasn't the money. It was the mechanism: verifiable on-chain attribution. Every trade tagged to its source, every payout automatic, every builder's contribution transparent.
Phantom didn't apply for a grant. They didn't pitch a committee. They built something people used and got paid for the activity they generated.
This is how developer incentives should have worked all along.
L1s and L2s have spent billions on grants. The playbook is familiar: announce a fund, open applications, convene reviewers, disperse capital, hope teams ship.
The problems are structural. Grants pay upfront for uncertain outcomes. Teams optimize for winning grants, not building products. Reviewers can't predict which projects will matter. And when the money runs out, builders leave because the incentive left with them.
Builder codes flip this. Instead of paying for promises, you pay for performance. Instead of one-time capital injections, you create continuous revenue streams. Instead of hoping builders stick around, you make staying economically rational.
Hyperliquid has distributed nearly $40 million to developers this way. Polymarket launched their builder program weeks ago. Kalshi just announced builder codes this week, alongside $2 million in grants for developers building on their new Solana integration. Base is implementing ERC-8021, a proposed standard that would bring this natively to Ethereum.
ERC-8021 proposes embedding attribution data directly into transaction calldata. A 16-byte suffix, a schema ID, and a reference to a registry contract where developers register their payout addresses.
The elegance is in what it enables: any application, any wallet, any agent can tag its transactions. Any protocol can decide how to reward that activity. The attribution layer becomes infrastructure, not something each platform builds from scratch.
Right now, implementations are siloed. Hyperliquid's builder codes only work on Hyperliquid. Polymarket's only work on Polymarket. Kalshi's only work on Kalshi.
ERC-8021 creates a shared standard. If multiple protocols adopt it, a developer's code could be recognized across all of them: one identity, multiple attribution sources.
Base's implementation, expected this month, will be the first major test.
Here's the interesting part about builder codes: they make incentive design legible.