The reflexes worth wearing in

A reflex is a move you’ve stopped narrating to yourself. This page collects the four that pay off most on a tool that lives in both a GUI editor and a terminal: posture, tool, model, and spend. None of them is new - each has a chapter of reasoning behind it, linked below - and that is the point. A reflex is what remains when the reasoning falls away and only the reach is left.

Choose the posture before you type

The highest-leverage reflex is the posture switch. Cursor’s built-in modes - Agent (the full-latitude mode that actually makes the change), Ask (read-only questions about your code), Plan, and Debug (for running down a tricky bug - its loop asks you to reproduce it) - are the same agent at different latitudes, and you switch them with the mode picker next to the chat input, Cmd/Ctrl+. for the mode menu, or Shift+Tab to cycle (shortcut reference). No single docs page lists the set (as of September 2026): Plan, Debug, and the Agent surface each have a page under cursor.com/docs, and Ask has no page there - its documentation is a Help Center article, ask mode, which describes it as a read-only mode for understanding your codebase. The reflex isn’t the keystroke. It’s reaching for the posture before you type, not after a run has already gone somewhere you didn’t want. That is a control decision in this course’s sense - control being what guides execution and enforces boundaries - and the boundary here is how much happens before you look.

The tell is the shape of the change in front of you. A contained fix in a file you own goes straight to Agent. A change that reaches past the file in front of you - a shared module, a schema, anything a half-dozen things depend on - gets Plan first: the mode that asks clarifying questions, researches read-only, and writes an editable markdown plan you approve before anything is built (the modes chapter owns the full judgment). Feel for that line before the first word, and the agent’s latitude matches the stakes without a deliberation each time.

Part of the same reflex is the two-model move from the models chapter: since Cursor 2.0 you can create a plan with one model and build it with another (changelog). When a task is worth planning, plan it with a frontier reasoning model and hand the written-down plan to a faster one to build - the plan is a markdown artifact, so the builder needs the plan, not the planner’s reasoning context.

Use the cheapest tool that fits

Tab, Agent chat, and @-pinning are three tools for three moments, and the reflex is taking the cheapest one that fits, without weighing them first.

• Tab - inline completion that predicts your next edit, accepted with the Tab key - is for the keystrokes you were already going to make. When you know exactly what you’re typing and just want it faster, you don’t open a chat; you let Tab finish the line.
• Agent chat is for the work you’d rather describe than type. When the outcome is clear but the keystrokes aren’t, you hand it a brief and review the diff.
• @-pinning is the exception, not the default. Since Cursor 2.0 the agent self-gathers context - the changelog’s words are that it “can now self-gather context without needing to manually attach it in the prompt input” (changelog) - so the reflex is to let it gather first, and reach for an @-mention only when you can see it guessed wrong: it’s reading the wrong file, missing the one that matters, or working a large repo where retrieval lands somewhere unhelpful. Pin to correct a miss, not as a ritual before every prompt.

Default the model to Auto

The model reflex, as the models chapter teaches it, is Auto by default, an override only when you can name why. Auto lets Cursor choose the model per request, and since July 2026 it can sit on the Cursor Router - the service Cursor uses to choose which model handles each request - with three optimization modes: Intelligence, Balance, and Cost (Router post). Hedge the mechanism before you lean on it, though: the Router modes are documented for Teams and Enterprise plans, and on other plans, including the free tier, Auto is simply the plain automatic choice (as of October 2026; models doc). What any request costs depends on your plan, routing, and current pricing, so the live picker and your usage dashboard are the source of truth, not a memorized rule. For the bulk of the work on any repo - a CRUD endpoint, a docstring, a fixture, a rename - Auto is the right first choice.

You break that default deliberately, not by accident. On the course’s illustrative budgetcli personal-finance API, the categorization rules engine - the one piece where a wrong design call is expensive to undo - is where you might hand-pick a stronger reasoning model, or want more context than the default. That second lever has a version history worth knowing: Max Mode, the context-window expansion older tutorials treat as universal, is legacy now. Cursor’s docs title the feature “Max Mode on legacy plans”: it is available only on legacy request-based plans, where it extends a model’s context beyond the default limit at the model’s API rate plus 20% (Max Mode docs, as of September 2026). The models documentation states the same boundary - “Max Mode is available only on legacy request-based plans” - so on current usage-based plans there is no Max toggle at all. What that doc documents about context is per-model - long-context limits and pricing, noted model by model - so on those plans the limit rides the model you pick, not a switch. The habit survives the change in surface: Auto for everyday, a stronger model where the reasoning genuinely pays, and back to Auto when the hard part is done. That is a capability-and-cost call in one line - capability bought only where it changes the answer, cost contained everywhere else.

Notice cost before a pattern scales

Billing is plan- and release-dependent. Treat the live Cursor pricing page and your usage dashboard as authoritative (as of October 2026); don’t assume that Auto, Composer (Cursor’s own model family), pinned models, or legacy Max map to one universal pool. Make expensive or long-context choices deliberately, and check the cost before repeating them - a choice that looks fine once is what a day of turns multiplies.

Try it

For your next three tasks, consciously pick the surface before you start: Tab for a rename, a bounded Cmd/Ctrl+K inline edit for a one-liner, Agent for a multi-file change. After each, note whether you picked the right rung or over/under-shot.

Those four reflexes settle your posture, your tool, your engine, and your spend. What they don’t settle is where a given task should run - and Cursor gives you four places to put it. Next: choosing the surface.