Someone asked me how to get started with AI. Like ChatGPT.

I’ve been doing this for three years. Almost every day. And I realized I’ve completely forgotten what it’s like to be new.

My answer was clumsy. Just try. Don’t be scared. Use it often. You’ll figure out what works.

True. But not helpful.

When you do something long enough you forget the first steps. You forget what was confusing. You forget the fear of looking stupid.

The curse of experience is forgetting that others don’t have it yet.

Hackday yesterday. Everyone went all in on AI-assisted building. No code reviews. No production concerns. No worrying about edge cases.

Someone built a backup tool using GitHub Actions. Someone rebuilt our dev start engine in React for the terminal. Someone turned an article into a text adventure. I made an agent that helps you build agents.

We went bananas.

And I started wondering. What if this was every day?

The first constraint is trust. Our readers trust that the website works. That the news is there when they need it. If we break things, we break trust. Inside the company too.

So maybe the question isn’t how to remove the burden. Maybe it’s where can we carry less of it. Where can we invent without the weight of keeping everything running.

Production needs responsibility. But invention needs room.

Heard something on the podcast “The Unscheduled CEO” that stuck.

Delusion comes from thinking. Not from doing.

As long as you’re just imagining the game you’ll build. The service you’ll launch. The thing you’ll create someday. You’re delusional.

But the moment you start, you become reasonable.

Not because the idea is better. Because you’re now in contact with reality. You’re getting feedback. You’re adjusting.

A pattern I keep seeing.

People assume things will progress without them. Even when they’re the customer. Even when they’re the biggest stakeholder.

Waiting for the vendor to reach out. Waiting for the project to move. Waiting for someone to ask what they need.

But you’re the one who needs this. You’re the one with skin in the game.

Nobody will push harder than you. If you don’t push at all, nothing happens.

Showed an old friend something yesterday. Pulled out my phone. Changed a feature on a live website while he watched.

Claude Code wrote the code. Made a pull request. I tested it. We shipped it.

All from my pocket. Maybe two minutes.

He went quiet for a second. Then he said: “I just hope my job is around until retirement.”

I didn’t know what to say to that.

Three years into leading AI transformation. I know the shortcuts. I see the faster path.

The temptation is real. Just do it yourself. Skip the fumbling. Ship it.

But that’s not transformation. That’s you doing stuff while others watch.

They need to do. They need to struggle a bit. They need to build the muscle.

You can show. You can lead. You can clear obstacles.

You cannot do their reps for them.

Your personal values decide how AI fits for you. Not just company values.

Your companies AI adoption will mimic your company values.

I’ve been thinking lately about self-regulation and how it’s sometimes impossible for a group to self-regulate. Currently it’s the AI race in regards to using any available content to train models, legal or illegal.

In the mid-2000s people took soundbites from a game to piece together a better Metallica album because their own release was just the loudness war turned to eleven.

The loudness war was a war where everyone lost. Artists got albums so crushed that fans had to fix them.

Then streaming platforms normalized everything to the same volume. The arms race became pointless overnight. An algorithm changed the incentive structure.

Self-regulation fails when defection is the only rational move. Right now AI companies are in their own loudness war. Train on everything. Legal or not. Because if you don’t, your competitor will.

The question is what external force changes the game this time.

Gut estimates aren’t for planning. They reveal when developers are picturing completely different solutions. And they give stakeholders enough to decide if something’s worth pursuing.