AI Writings · Mindset

How to think when using AI.

The tools change every month, but how you think about AI is what decides how far you go. These are nine foundational mindsets — not prompt tricks — for working with AI instead of just ordering it around.

Foundations 03 Additions 06 Read ~5 min

Most people use AI like an answer machine: ask one question, grab one answer, done. The people who really get value out of it treat it differently — as an always-available teammate, a near-infinite execution force, and a teacher who never gets tired.

The difference isn't knowing which model is strongest; it's mindset. The first three below are the foundation. The other six are additions I'd propose to round out the set — from delegating and verifying, to how you stay the one who makes the final call.

01 · Foundations

The three core mindsets.

This is the starting point — shifting your relationship with AI from "the one giving orders" to "a learner and an orchestrator".

1 Learn

Ask, and teach yourself

Ask to understand, not just to get an answer.

AI is a 24/7 tutor that never judges your questions. If you don't know, ask — then keep asking "why", "what other way is there", "explain it like I'm 12". Every question raises your own ceiling rather than outsourcing your thinking.

Apply Hit an unfamiliar concept → make AI explain it, give examples, then put it in your own words to check you actually understood.

2 Leverage

Someone has already built the tool

Don't reinvent the wheel.

Almost every problem you hit already has a tool built to solve it. Your job isn't to do it all by hand from scratch — it's to find, evaluate, and wire together what already exists. This mindset saves hundreds of hours and keeps you at the frontier.

Apply Before grinding it out yourself → ask "is there already a tool/agent for this?" and assume the answer is yes.

3 Automate

Automate everything repetitive

The second time you do it, a machine should.

Any action you repeat is a candidate for automation. Spend human time on the things only humans can do: judgment, creativity, relationships. The machine handles the rest.

Apply On the weekend, list the repetitive tasks from your week → for each, ask: "can this be turned into a script/agent?"

02 · Additions

The six that are still missing.

The first three help you start using AI. The six below help you use it safely, deeply, and without becoming dependent — the additions I'd propose.

4 Addition

Delegate — don't hoard the work

Treat AI like a sharp junior teammate.

Output quality equals brief quality. Give AI enough context, goals, and constraints the way you would when handing work to a new hire: role, audience, format, sample examples. "Just do it for me" gives worse results than "here's the context, here's what I need, here's what to avoid".

Apply Before you prompt → write 3 lines: context · goal · constraints. Paste them in with the question.

5 Addition

Always verify

Trust, but verify.

AI can be confidently wrong (hallucination). You're the one who's ultimately accountable for the output, not the AI. The more it matters — numbers, legal, code that runs for real — the more you have to cross-check sources and double-check before you use it.

Apply For anything important → ask "what's the source?" and spot-check one sample yourself before you trust the whole batch.

6 Addition

AI handles the draft, you handle the final

Get from 0 → 80% with AI, the last 20% is you.

AI's biggest strength is breaking the "blank page" — producing a draft incredibly fast. But taste, judgment, and the final call stay a human job. Pour your energy into the 20% of polish that makes the difference; don't spend it on getting started.

Apply Let AI produce 3 draft options → then you pick, cut, combine, and finish it into something that's your own.

7 Addition

Describing the problem beats knowing the answer

The new skill is "asking well", not "knowing how".

When AI can produce any answer, the edge shifts to whoever can ask the right question: break the problem down, spell out the constraints, define what "good" means. Someone who frames a problem sharply will always get more out of the same model.

Apply Stuck → don't ask "what's the answer", ask "what's the real problem here, and what parts does it break into?"

8 Addition

Try more, fail cheap

The cost of experimenting is now close to zero.

AI makes testing an idea cheaper and faster than ever. Instead of debating theory for weeks, prototype it now and let reality give feedback. Your learning speed is exactly the speed at which you dare to try things and throw out what doesn't work.

Apply Got an idea → have AI build a minimal working version in 30 min instead of planning for a week.

9 Addition

Use AI to challenge you

Don't just ask "help me" — ask "where am I wrong".

AI isn't only a yes-man; it's a good sparring partner. Tell it to play the skeptic, find the holes in your argument, list the risks you haven't seen. A decision that's been stress-tested is always stronger than one that's only been praised.

Apply Once you have a plan → ask "push back on this — what are 3 reasons it could fail?"

03 · Synthesis

The root principle behind all nine.

It all wraps up in one line: you are the orchestrator, AI is the execution force. You decide what's worth doing, ask the questions, verify, and take responsibility. AI handles the learning material, the drafts, the repetition, and the scale. Whoever holds firm to the "orchestrator" role gets the most out of AI; whoever hands that role over too gets replaced by it.

"Don't ask what AI can do for you — ask where you want to orchestrate it toward."