You reason through a tricky architecture question in Ask, then switch that chat to Agent and say "now implement what we just discussed". How much of the discussion does Agent have?

The principle — Each mode uses its own context, so switching modes starts a fresh context window. Cursor advises starting a new chat when you change tasks.

Why the key is correct — The reasoning you did in Ask stays in Ask. An instruction to implement what you just discussed refers to nothing the agent can see, so it fills the gap with a plan of its own.

Why the others are wrong — Modes are more than permission levels, because context is scoped per mode too. No seed message is handed across; the transcript still on screen is display, not context. Nothing is summarised on the way over, which is why the result can arrive confidently wrong.

Remember this — A mode switch is a fresh start, so restate the conclusion in writing once you have switched.

Sources — Cursor's Agent mode help documentation.