What’s the messiest part of the AI prototype feedback loop right now?

Curious to hear how others are currently handling review and feedback for AI-generated apps, prototypes, dashboards or websites.

The generation side of AI workflows has accelerated incredibly fast over the past 12–18 months.

But once something gets created in Claude, Cursor, Codex, Bolt etc… the collaboration and review process still often feels fragmented.

Are people mostly using:

• screenshots?

• Loom videos?

• Slack/Teams threads?

• staging environments?

• Figma?

• GitHub?

• Live links?

And where does the workflow start to break down?

For us, one of the biggest friction points has been contextual review and keeping feedback connected directly to the artefact itself as iterations happen.