Automatic code review on every PR has a human bottleneck. With maintainers and developers driving agents to do more and more; maintaining quality code is harder than ever which in turn makes maintaining in general more challenging. Without the ability to automatically enforce code conventions in a deterministic way, non-deterministic models are used (often in waves/layers/fleets) in attempt to verify all sorts of "things" such as meeting the contributor requirements, following local checks, cleaning up docs, committing in a certain way, etc. This is an inefficient solution that is not able to provide 100% certainty that something was checked. Exploring the combination of deterministic checks, Decision Models, and LLMs means when it's time to take a look at a PR it has a lot more relevant available. > Below are a few screenshots of Corbits Triage overlayed directly on GitHub and with the extension open

Gave my agent $1 and access to @nansen_ai api via @lanatools. It spent 20 cents and came back with 3 Solana tokens that smart wallets are buying. Kinda wild that research costs 20 cents now. Lanatools - connecting your agents to @solana.