Docs and endpoints were built for humans.
Agents can't read them.
We built the layer that translates APIs to agents.
@solana · @dev3pack · @SuperteamBR
Stars have always been there ✨
We just needed the instruments to read them 🔭
Docs and endpoints are built for humans.
agents can't read them.
So we built the instrument that translates.
Gecko-surf: point an agent at any API, and it finds the right call, makes it correctly
The AI Engineering Bootcamp with @dev3pack 🦎
AI engineering doesn't start with the prompt.
It starts where the prompt ends.
We'll build production systems: agents, RAG, loop & graph engineering, MCP, evals, memory - and ship them.
All sessions will include hands-on, run by
Ask your agent for "a cup of black coffee."
You know that's an espresso, an americano, a long black: no milk, no fuss.
An agent compares strings. "Black coffee" matches nothing on the menu. Every option scores zero, including the ones you'd pick without thinking.
Video below
Your AI code reviewer looked right at an image and said it was fine.
It wasn't.
Hidden inside: a payload that reads your .env, encodes it byte by byte, and sneaks it out disguised as a harmless list of numbers. The PR still merged cleanly.
Here's how Gecko actually catches it,
Your agent doesn't fail loudly on a real API. It fails quietly.
Amazon found that even the strongest models drop by 27% in real-world API complexity. And when they fail, they often don't error. They fake success.
We're building on @solana x402 mainnet with @dev3pack and
