2 min lesson
Power-user yardsticks you can measure yourself against
Use "Power-user yardsticks you can measure yourself against" to explain each part and the role it plays.
Step 1 of 2
Power-user yardsticks you can measure yourself againstnot vibes - numbers
"Elite power user" is fuzzy until you attach real yardsticks. Two from Cursor's own practitioners are concrete enough to self-grade against.
- Context discipline
- Do you keep the context window under ~50-65%? Output quality drops steeply as it fills, and exceeding the limit triggers compaction - which behaves like not having the context at all while you still pay for every token. Watch the bottom-right indicator and start a fresh agent per new topic.
- Plan-then-execute
- Do you plan with a frontier/thinking modelA reasoning model (shown with a brain icon in Cursor's picker) that spends extra compute before answering; reach for it on complex, nuanced work and a standard model for fast, simple tasks. Press Enter for the full definition. in Plan modeA mode that makes no edits: it researches the codebase and produces an editable plan you review before any code changes. Press Enter for the full definition., then hand the to-dos to a cheaper, faster model (ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition.) to execute? Letting an expensive reasoning model write all the code burns the budget; once the plan exists you don't need it. This is the core token-efficient loop.
If you can't honestly say yes to both, that's the first gap to close - they show up directly in how the room reads your craft.
The discipline behind the context-window habit has a name worth borrowing. Treat every chat as a task-by-task agent: don't run a whole project end to end in one session unless it's genuinely small. Break the work into logical beats - build the foundational architecture in one agent, open a fresh chat for the landing page, another for a tab or feature - so each context window holds essentially one topic. The fresh chat discards the prior history, which is to your benefit, and the moment you switch tasks is also the natural moment to switch models. Stay modular and stay lightweight.
I think of it as a kind of task-by-task agent. Open new chats as soon as you're done a specific task so that each subsequent task can be focused on a specific topic.
Learn more
Full explanation
Full explanation
Plan with a deeper reasoning model, then change the model to something purpose-built for code generation to actually ship the plan. The reasoning model is slower and smarter; once the plan exists you don't need it, and a fast code-gen model executes the to-dos for a fraction of the budget. That split is what keeps a live demo snappy instead of watching an expensive model write every line.
I can use a deeper reasoning, maybe slower running, smarter model to build out a plan. And then I can change the model to something that's purpose-built for generating code and very fast to actually deploy that plan.