Lesson 1.4 · Tier 1 — Prompting
Here’s a failure you’ve already experienced without naming it: you ask for “a great Instagram post about my new product” and receive something… fine. Competent, complete, forgettable. The diagnosis isn’t a missing bone from Lesson 1.0 — it’s that you asked one prompt to do two jobs: decide what’s worth saying and say it well. Forced to do both at once, the model compromises on both. The best angle never got explored, and the writing hedged toward average.
Separate the jobs and both transform.
Thinking mode: ban the output
In thinking mode you want options, questions, and disagreement — explicitly not the deliverable. The key phrase, which feels strange the first time: “Don’t write it yet.”
Notice what this buys: seven directions instead of one silent default, plus a reasoned recommendation you can argue with. Thinking mode is also where the personas from Lesson 1.2 live — the coach and devil’s advocate are thinking-mode tools by nature.
Producing mode: lock the spec
Once the thinking has picked a winner, flip modes completely. Now ambiguity is the enemy and the four bones go in hard:
Producing mode wants no creativity about what — only excellence in how. Every choice you leave open here is a choice made by the average of the internet.
The tell, and the workflow
How do you know which mode a task needs? One question: “Do I already know exactly what good looks like?” Yes → producing. No → thinking first, always. The expensive mistake is skipping straight to producing to feel fast — you save two minutes and marry the first idea the machine happened to average into existence.
For anything that matters, run the two-step: a thinking exchange to choose the direction (often worth doing in its own chat, per Lesson 1.3), then a locked producing prompt, then the pushes from Lesson 1.1. That sequence — think, lock, produce, push — is the complete Tier 1 skill assembled, and it’s exactly how you’ll build things in Tier 2.
Take one real task you’d normally one-shot and run the two-step: thinking prompt (with “don’t write it yet”), then producing prompt. Compare against what one-shotting would have given you. Save your thinking-mode opener as library entry #5.
Last lesson of the tier: what AI can make beyond words.Lesson 1.5 — Beyond Text →