docs: simple chat UI analysis

[mistercrunch]

Summary

Analysis doc at docs/internal/simple-chat-ui-analysis-2026-05-08.md answering the question "Should Agor ship a simple, ChatGPT-like entry point for sessions?"

Recommendation: Don't build a separate simple chat UI. The four core arguments:

1. Wedge. The positioning doc says Agor exists because single-chat is the wrong primitive at team scale. A chat entry point inverts the message at the door.
2. Friction diagnosis is wrong. Onboarding drop-off is BYO-runtime credentials + repo, not "the canvas is scary." A simpler UI doesn't fix what's actually broken.
3. Schema doesn't support it cheaply. Sessions require a worktree FK end-to-end. Either you fake one (hide chrome on existing UI = focus mode) or you fork the data model (compounding tax forever).
4. Pattern match cuts the other way. Linear Asks, Notion AI, Slack-Claude all live inside the existing UI — none of them shipped a sibling app.

The document reframes the question into four variants (alt-app, focus mode, public share, Scratch) and recommends building focus-mode + Scratch (already brainstorm doc #12) and holding public share until asked.

What's in the doc

• Reframing of the question into four distinct variants
• Steel-man for and against (eight against-points)
• Frameworks: wedge, 2-year vision, funnel diagnostic, MVP variants ranked by reuse, reuse-not-duplicate alternative
• Committed recommendation
• "What I'd want to know to be more confident" — funnel data is the load-bearing missing input
• Open follow-ups (incl. whether /s/:sessionId is already implicitly there)

Draft PR — analysis only, no code, not for merge until reviewed.

Test plan

• Read the doc end-to-end and push back on anything that's wrong
• Confirm the recommendation matches strategic intent (or argue why not)
• Decide: do focus-mode + Scratch land on a roadmap, or stay in the brainstorm pile?

🤖 Generated with Claude Code