Manifesto

Most code is now written by agents, and that share only grows. A codebase optimized for a human reader is the wrong shape for the reader that actually writes it. Porto AI is built on the opposite premise: design the architecture for the agent first, and let humans steer.

We believe:

• The author has changed. Optimize the codebase for the reader that writes most of it, not the one reviewing it.
• One reader, not two. You cannot optimize a single codebase for humans and agents at once. Pick the agent, and commit to it.
• Tokens are the budget. Fewer files, fewer hops, less indirection. Every traversal and every clarification round has a cost.
• Density over decoration. Code should be explicit and unambiguous for a machine, not softened for human reading.
• Autopilot is the goal. Autonomous feature work, debugging, and maintenance with minimal human steering.

This is not version 2 of Porto SAP. It is a separate architecture for teams who let agents own the codebase.

The full specification, the structure, the conventions, and the way it all fits together, is in private testing and shared directly. See Early Access.