refactor: improved flow in ai doc

Author: devvaannsh

docs/03-ai.md

@@ -27,7 +27,7 @@ Free users get a daily and monthly chat limit. Once you're past halfway on eithe
 - **[Plan, Auto-Edit, and Edit modes](#permission-modes)** let you decide how much freedom the AI has. Plan mode confirms before any change, Auto-Edit applies safe edits automatically, and Edit lets you approve each action.
 - **[Restore points and a visual undo timeline](#undo-and-restore)** mean every AI change is reversible. If you don't like what it did, roll back with one click.
 - **Live preview integration** — the AI can see your running app, take screenshots, click around, and verify its own work.
-- **[Choose the model](#choosing-a-model)** you want for each chat, even in the middle of a chat.
+- **[Choose the model](#choosing-a-model)** you want, and switch it even in the middle of a chat.
 - **[Bring your own provider](#settings)** — Claude from Anthropic (the default) gives the best results, or bring your own API key from any Claude Code CLI-compatible provider.
 - **Privacy-first onboarding** with a clear consent dialog and a video walkthrough.
 - **[Type while the AI is still working](#sending-messages)** — your next message gets queued.