AI Pattern Explorer
Pick an AI design pattern and see how it works at a glance, its flow, strengths, watch-outs, and an example prompt. The patterns behind modern agents, side by side.
The patterns behind modern agents
Building with AI is settling into a small set of recurring shapes, the same way classic software did with design patterns. This explorer puts them side by side: select a pattern to see its flow, what it’s good and bad at, when to reach for it, and a prompt you can adapt. Each links to a full write-up in the AI Patterns Library.
The starter set
- Reflection: the model critiques and improves its own output.
- Router: classify the request and route it to the right path.
- Judge: one model evaluates another’s output against a rubric.
- Verification: check the claims before responding.
- Human-in-the-Loop: request approval before acting.
FAQ
Are these official patterns?
They’re the names the community has converged on across agent frameworks and provider docs. The value is collecting them in one place with honest trade-offs, not coining a standard.
Can I combine patterns?
Almost always, real systems chain them. A Router picks a path, that path uses Reflection, and a Judge gates the result. The patterns are building blocks, not either/or choices.
Is anything uploaded?
No, the explorer is static content rendered in your browser. Nothing is sent anywhere.
Related
Read the full AI Patterns Library or the concepts in AI Explained.