@yiyangleex: Using tools like Codex and Cursor to learn and understand codebases has a major issue: if you want to deeply understand a project's architecture, they can often only provide you with a bunch of markdown text or Mermaid diagrams.

Using tools like Codex and Cursor to learn and understand codebases, the biggest problem is: when you want to deeply understand a project’s architecture, all they can give you is often a bunch of markdown text or Mermaid diagrams.

Passively reading text and staring at static diagrams to guess the structure—frankly, it’s quite anti-human.

In modus, AI can directly generate interactive diagrams, schematics, and even sandbox HTML inline in conversations—you can drag, click, and physically “touch” these concepts, instead of staring blankly at a diagram and trying to imagine it yourself.

This might be the most natural experience in AI interactive visualization so far.

Below is a test I did using the Codex CLI source code repository, letting GPT 5.6-Sol in modus explain the agent loop principle using inline interactive diagrams. The entire execution chain became clear at a glance—much more intuitive than spending half an hour reading documentation.