I’m Adrean. Fido started while I was building a Godot deck-builder with AI. New chats wrote code fast and forgot the game. I kept re-explaining mechanics or shipping the design-wrong thing.

The hardest work wasn’t the parser. It was deciding what Fido should be for.

I first bet on an audit report: mark the GDD, scan GDScript, catch drift later. Honest, but late. With ChatGPT I pressure-tested that choice against what actually hurt every fresh session had lost the design. So I flipped it: context first, audit second. Same local engine; the hero path became refreshing a short Game Design Context block before coding. The report stayed for when I wanted proof.

Other calls were just as sharp. Local-only scans no uploading design or source. Explicit markers as the only authoritative coverage, so suggestions wouldn’t fake precision. Codex/ChatGPT first as the ship surface, not boiling the ocean on every host. And a standalone ZIP, because “clone the monorepo” wasn’t a real install.

Those decisions are what Fido is. It doesn’t replace your judgment it puts your design in the room before the AI starts.

Built With

  • cloudflare-workers
  • codex
  • gdscript
  • godot-4
  • gpt-5.6
  • python
  • r2
  • react
  • vite
Share this project:

Updates