GuideOct 18, 20264 min read

The prompt I run before any long AI build session

The single biggest jump in my output came from planning before the model starts. Here is the two-prompt routine, free to copy.

When a new, more capable model came out this summer, I moved four or five projects past the idea stage in a week. The model was not the main reason. The planning was.

Prompt one: prove you understood me

I record a voice note about where everything stands, paste the transcript, and tell the model its first job is not to solve anything. It has to reconstruct my context: projects, constraints, decisions, blind spots. Only then does it suggest an order of work.

Prompt two: run it end to end

Then each project gets a brief with an objective, context, priorities and one execution rule: run this end to end, and only stop for decisions that are irreversible or truly mine. The model finishes, then lists the decisions it made on my behalf.

  1. Voice note: where everything stands (10 minutes).
  2. Context capture prompt: the model explains it back.
  3. Pick 1 to 3 tasks for the session.
  4. One end-to-end brief per task.
  5. Save the outputs so the next brief builds on them.
Tools change. A good brief works on every model.

All three prompts are in the library with copy buttons: context capture, the end-to-end brief, and the session planner for when you have more projects than hours.

Get next week's update.

One workflow, one number, one giveaway, every week. Unsubscribe anytime.

One email a week at most. Unsubscribe anytime.

Want this in your business, not just on your screen?

Book the free audit: one hour, a one-page map in 48 hours.

Book your free AI audit