patrickz.aiLet’s talk ↗Field notes
Each field note teaches one idea about using AI well, told through the real project where I learned it. Follow the path in order or jump to what you need. Every idea ends with a hands-on guide so you can try it the same day.
Stage 1 of 5
What AI is actually doing, what it needs from you, and how to ask for the right thing.
Start hereIdea 01 · 8 min readRead ✓
What a chat assistant is really doing when it answers you, why it can sound certain and still be wrong, and a simple habit for trusting it the right amount.
Idea 02 · 8 min readRead ✓
Learn what an AI can actually see when it answers, and how to hand it a small, dated context pack that turns a generic reply into one that fits your life.
Idea 03 · 7 min readRead ✓
Many disappointing AI results answer the wrong question. Here's how to pin down the real problem first, so the AI's speed works for you.
Stage 2 of 5
Judge the output yourself and choose tools on evidence, not rankings.
Idea 04 · 7 min readRead ✓
Why making more with AI won't improve your results, and a simple practice loop that trains the judgment that will.
Idea 05 · 8 min readRead ✓
Why AI rankings can't choose your tools, and how to test them on a small bench of your own tasks.
Stage 3 of 5
Have AI write and review against your checklist, and use it to check your own patterns.
Idea 06 · 8 min readRead ✓
Learn the create, review, revise loop: AI writes better instructions, checks its draft against your checklist and fixes one thing at a time, and you make the call.
Idea 07 · 8 min readRead ✓
Learn how to let AI find patterns in your own notes, calendar or chats, make it show its evidence, and turn one finding into a small experiment you can test.
Idea 08 · 10 min readRead ✓
A getting-started guide and a five-minute video for Microsoft 365 Copilot Cowork: have it find a repetitive process in your email, map the decisions, turn them into a shareable skill, and run that skill on a schedule, with guardrails that keep you in charge.
Stage 4 of 5
Ship small software with an agent, and fix it calmly when it breaks.
Idea 09 · 8 min readRead ✓
Learn to tell an AI coding agent what “done” looks like before it builds, then ship a small app in thin slices you can prove actually work.
Idea 10 · 8 min readRead ✓
Learn six steps for finding out why something broke (a game, a spreadsheet, a prompt) so the AI works from evidence, not guesses.
Stage 5 of 5
Direct several tools at once, and build for one real person.
Idea 11 · 12 min readRead ✓
Space wizards. Expiring credits. A production workflow that became almost as interesting as the film.
Idea 12 · 12 min readRead ✓
One hundred and four posts, two years of ideas and zero websites. Here is the exact workflow, the prompts and the guardrails that turned them into patrickz.ai, plus the instructional video I made about it.
Idea 13 · 13 min readRead ✓
A request to train 20 new team members became a practical test of GPT-6: could the workflow I used for anime music videos make useful training material while I slept?
Idea 14 · 10 min readRead ✓
In one day, parallel AI agents rebuilt Super Mario Bros. and The Legend of Zelda starring me, upscaled and over the top, playable on my phone. The real lesson was how to direct agents and let a bot judge their work.
Idea 15 · 8 min readRead ✓
Learn to start an AI project from one person's real needs, pick the easiest input that works for them, keep a non-AI fallback, and measure what matters to them.
Want to practice first? Browse all 57 hands-on guides ↗Get new field notes by RSS or follow Patrick on LinkedIn ↗.