What happens when an AI architect performs 10 reviews of their architecture?

Source: Dev.to

Introduction

I launched arch‑review today — an open‑source architecture review assistant with four specialized agents running in parallel. The feature I’m most proud of isn’t the speed; it’s the memory.

Evolution Cycle

After each squad review, each agent logs a lesson. For example:

🔐 Security Agent: “Always check for rate limiting on the API Gateway before checking internal authentication.”

These lessons are injected into each agent’s prompt for the next review. Over time, the squad calibrates to your tech stack, your failure modes, and your business context.

Evolution Levels of Each Agent

The Evolution Dashboard tab shows this information in real time:

• Total revisions
• Lessons per agent
• Cross‑patterns discovered
• Findings accumulated over time

What’s Included in the Project

🔗

How to Use

If you’re an architect, tech lead, or senior engineer, work on an architecture you know well. You’ll see things the team hasn’t seen. After five reviews, the squad will know your architecture better than any tool you’ve ever used.

Support

⭐ If the tool makes sense for your work, starring the repository helps a lot in reaching more people.

Note

The squad’s loading message says:

🎸 The cost agent is playing the budget blues. What a sad song.

Because waiting should have personality. 🎭