Here is the trick almost nobody uses: do not start with your weaknesses. Start with your strengths.
Most people open an AI tool and reach for the things they struggle with. The logic seems right — AI fills gaps. The problem is you cannot evaluate what you cannot already judge. When AI produces something in a domain you barely understand, you have no way to know if it is good. You accept it and learn nothing.
The people getting genuinely skilled at AI do the opposite. They take a task they already do well, do it their way first, then give the same task to AI — and study the gap.
Where AI agrees with you: validation. Where it goes a different direction: either it has a blind spot, or you do. Where it does something you would never have thought of: a new technique to examine.
That gap is the curriculum. Run this ten times in your domain and you will understand how AI actually thinks — not how the marketing says it thinks.
There is a second-order benefit most people miss: you learn your own mental models better. AI forces you to articulate why it is wrong. Your tacit knowledge becomes visible.
📚 THE READ
The Expert's Advantage in the Age of AI — Why subject matter expertise is becoming more valuable as AI gets better, not less. The people who can evaluate AI output are the ones building the most leverage.
Deliberate Practice in the AI Era — Improvement requires feedback loops. AI can be the fastest one you have ever had — if you set it up correctly.
What AI Gets Wrong About Your Industry — Every field has systematic gaps in AI output. Most people never notice because they never test against their own work.
🛠 THE TOOL
NotebookLM (Google) — Upload documents from your domain and it builds a model grounded in your actual source material. You control the inputs, so you can evaluate the outputs against your own expertise. Free with a Google login.
💡 THE PROMPT
Use this to run the comparison exercise:
I want to compare approaches on [task]. Here is the brief: [brief].
Work through this independently. Commit to a direction — do not hedge. Give me your complete version first.
After you share, I will share mine. Then we compare: where we agreed, where we diverged, and why. No clarifying questions yet.
Forcing a committed first attempt gives you a real comparison point instead of a negotiated one.
🎯 THE ASK
What is the one domain where you have genuine expertise — and have you ever actually tested AI against your own work in it?
Reply directly to this email. I read every response.
If this method resonates, the next step is building it into a system you actually run every week — not a one-off experiment. That is exactly what the Power Guides at tminusai.com are built for: a personal AI operating system with the prompts, workflows, and project structure that make this repeatable.
There is also a free GPT-5 Playbook on the site if you want to start without spending anything.
More next week.
Kapish