gblin (gblin)

Honest question for solo devs: who reviews your code? Mine gets read by three AIs, but each one first has to pass a reading test on the file. If it can't answer questions whose answers are written in the code, I don't trust its findings. How do you handle it?

• 0 replies
• 8 recasts
• 15 reactions

What one person and three AI reviewers shipped in a year on Base: - a basket vault (cbBTC, WETH, USDC) minted and redeemed at NAV - a rebalancing auction that CoW Protocol solvers already fill - a crash shield that cuts the weight of an asset that crashes - a redemption nobody can pause - 591 tests, fork tests, mutation testing and symbolic execution, all written up in public No team, no funding. Just a lot of evenings.

• 0 replies
• 5 recasts
• 8 reactions

Before building the treasury plan I read how the big ones handle idle cash. The pattern is always the same: operating cash stays in USDC, the surplus goes somewhere, and it comes back just in time when a payment is due. So I didn't invent a model. I built the missing step: one call that shows the numbers before anything is signed, including what it costs to get out.

• 0 replies
• 3 recasts
• 3 reactions

No team, no funding, no marketing budget. One person on Base, with three AIs as tireless reviewers, trying to build something that holds up next to what funded teams ship. That's GBLIN: one token holding cbBTC, WETH and USDC, minted and redeemed at NAV straight from the contract. When the weights drift it doesn't trade on its own: it opens a Dutch auction anyone can fill, CoW Protocol solvers included. Parameters only move through a 48h timelock. No paid audit yet, and I say it upfront. What I have instead: 591 tests, 132 fork tests against the live pools, 6 hours of coverage-guided fuzzing, mutation testing, and every line read by three AIs. The full review record is public. That's the challenge. Tell me where it breaks: github.com/gblinproject/GBLIN-Protocol

• 0 replies
• 2 recasts
• 3 reactions