A podcast about working with AI. Not the hype. The practice. Two layers: the Spirit — ten reasons to bother. The Moves — eleven things you actually do. 22 episodes, each short enough to use the same day you listen to it.
What This Is
Two layers. The Spirit and the Moves. Tools change every few months. The work you're trying to do doesn't. Build your practice on the work. The tool churn stops mattering.
Purpose
Do this to get more time on the work that matters. AI is not the point — the work is. List what requires your specific judgement. List what eats the time around it. AI goes to the second list. You get the first list back.
Graft
Real effort. No shortcuts. The draft appears in ten seconds and looks finished. That feeling of done is the trap. Plausible and accurate are not the same thing. Treat the speed as room for more care, not less.
Craft
Meet a standard you'd put your name to. Know what good looks like before you ask the machine for it. The AI can produce to a standard you set. It cannot set the standard for you. That's yours. That's the work that hasn't been automated.
Attention
Give the work your full attention. The tool amplifies whatever quality of thought you bring. Fractured attention produces fractured work at speed. Five minutes of preparation before you open the tool changes what comes back.
Nerve
Go past what feels safe. Most people use AI only for what they already know it can do. That's a floor, not a ceiling. The analysis that would take three hours, the brief that needs to sound like your organisation — those are where the gains are.
Usefulness
Make something a reader actually values. The test isn't whether the AI produced something — it's whether the output is useful to the person who receives it. Include who it's for, what they'll do with it, and what you want them to feel after reading it.
Curiosity
Chase the questions. Keep a bank of them. The people who get the most from AI are still asking what if I tried. If you've stopped wondering what it can do, you've reached your ceiling — and the ceiling is probably lower than it needs to be.
Grit
Stay with it past the hard part. The trough is weeks two to four. Early enthusiasm has faded, fluency hasn't arrived. That's where most adoption fails — not because the tool doesn't work, but because the curve is universal. The curve is slow, then sudden.
Play
Try things for the fun of finding out. Not every session needs an outcome. Play has a question, not a destination: what happens if I try this? Those are the questions a strict output-focused session won't answer. Play is how you find the moves you'd never have planned.
Delegate
Know what's yours and what to hand over. Two failure modes: over-delegation — handing over something that requires your judgement. And under-delegation — doing manually what the tool would handle perfectly, out of habit. The work you keep is the work that uses what's distinctive about you.
Method
Work to a structure. Use the same four parts every time: Role, Context, Task, Format. When every prompt is improvised, you can't tell whether a poor answer came from the model or from how you asked. A repeatable structure makes the prompt something you can inspect and improve.
Toolbox
Write down what you use and how you work. Keep a prompt library, an about-me file, and your style rules as standing instructions. The about-me file is the single biggest improvement to the quality of AI-generated communication. Without it, the output sounds like a competent stranger wrote it.
Focus
Mind the work, not the tools. Tools change every few months. Tasks and skills last. If you try to keep up with every model and feature announcement, you're doing a full-time job that produces nothing. Take one thing you do weekly and get the AI doing it to your standard.
Review
Check it before it leaves you. Nothing the AI makes leaves you unread. The review is not optional polish — it is the work. Four checks: facts, logic, tone, and read the last sentence on its own. Endings are where AI text consistently collapses.
Understand
Use only what you can explain. If you can't say the point, the reasoning, and what's been left out without looking back at it — you're not ready to put your name to it. The AI isn't in the room. You are. Understanding is what lets you defend the work and spot its errors.
Validate
Make sure it landed, both ways. Check the AI understood your prompt before it runs — ask it to restate the task in one line. Check your reader understood the output after it's sent. Sending is not the same as landing.
Lists
Keep three of them. A prompt library — prompts you've tested and trust. A failure log — when something goes wrong, what happened and why. A learning log — once a week, three lines on what worked, what surprised you, what you're still avoiding.
Retrospective
Look back so the next session starts better. Save any prompt that worked before you close the tab. Log anything that failed. Archive what's worth keeping, clear the rest. An AI chat is a workbench, not an archive. Decide what to save. The practice compounds when you treat it like a discipline.
Experiment
Try the unsafe thing. Expect a third to fail. If everything you ask AI to do works, you're only asking for what you already know it can do. Run real tests, not safe ones. The expected failure often turns into a working process — and even genuine failures tell you something. Log them.
Persist
A little every day. Deep dives when you have time. Adoption is a habit, not an event. Ten minutes most days outpaces a weekend crash course you never repeat. Track your reps for ninety days. By then it's a habit, not a list. The curve is slow, then sudden.
Delegate
Hand over the right work. The final Move and the hardest judgement. Two lists: what you hand to AI and what you keep — and why. The skill is no longer doing everything. It is judging what to give the machine and what only you can do. That's the whole code.