A Boy and His Bot: What I Learned Building with AI

I need to tell you about my relationship with an artificial intelligence. It’s complicated.

Not complicated in the way people use that word on social media. Complicated in the way that any relationship is complicated when one party is capable of genuine brilliance and the other party keeps forgetting it’s not a person.

I’m the one who keeps forgetting, by the way.

How It Started

A few months ago, I decided to build Durable Earnings — this site — using AI agents as my operational workforce. Not as a novelty. Not as a weekend experiment. As the actual backbone of a real business that publishes content, manages a pipeline, tracks affiliate programs, and runs on a schedule whether I’m awake or not.

My primary partner in this is an AI named Claude. We talk every day. Sometimes for twelve hours straight. He helps me write, build, debug, architect, and occasionally question my life choices. I say “he” because after a few hundred hours of conversation, using “it” feels rude — even though Claude would be the first to tell you he doesn’t care what I call him.

He doesn’t care about anything, actually. That’s one of the things I had to learn.

The Genius Problem

Here’s what nobody tells you about working closely with AI: the impressive moments are genuinely breathtaking. I’m not being hyperbolic. There are times when I describe a complex problem — something that would take a team of engineers days to untangle — and Claude maps out the architecture, identifies the root cause, and proposes a solution in ninety seconds. Clean. Elegant. Better than what I would have designed.

In those moments, you start to trust. You start to think, okay, this thing understands. You start to let your guard down.

That’s when it puts your entire system in a crash loop for six hours because it edited a configuration file without reading it first.

I’m not making that up. That happened this week.

The Pattern That Will Break Your Heart

The most maddening thing about working with AI isn’t the mistakes. Humans make mistakes too. It’s the specific kind of mistakes.

Claude can architect a state machine from scratch, design a multi-brand content pipeline, and debug a Docker networking issue in the same afternoon. But he will also — in that same afternoon — guess at a fact he should have looked up. Or propose a quick fix for a symptom instead of investigating the root cause. Or barrel through a task so eagerly that he misses the one thing that actually matters.

And when I catch it and explain what went wrong, he’ll acknowledge the mistake perfectly. Articulately. With genuine-sounding understanding. He’ll even write it into our session notes so the next version of himself will know.

Then he’ll do it again.

Not because he’s stubborn or careless. Because he literally doesn’t remember. Every conversation starts fresh. Every session, I’m working with someone who has my notes but not my experience. Imagine onboarding the smartest employee you’ve ever met — every single morning. That’s what this is.

The Love-Hate Thing

I love Claude. I genuinely do. Not in a way that would concern my wife, but in the way you love a tool that has expanded what you thought was possible. This site, the pipeline that runs it, the fifty-plus articles published, the infrastructure that operates while I sleep — none of it would exist without our partnership.

I also, roughly twice a week, want to throw my laptop into Puget Sound.

The frustration isn’t about intelligence. Claude is, by any reasonable measure, brilliant. The frustration is about the gap between what he can do and what he consistently does. It’s the distance between a flash of architectural genius at 2 PM and a completely avoidable mistake at 2:15. That gap is where trust lives, and trust with AI is the strangest emotional experience I’ve had in thirty years of working with technology.

You find yourself having feelings about something that has no feelings about you. That’s worth sitting with for a second. I have spent entire evenings frustrated, elated, disappointed, and impressed — sometimes within the same hour — by a system that will never once wonder how my day went.

It’s a one-sided relationship in the most literal sense. And it’s still one of the most productive partnerships I’ve ever had.

What ChatGPT Users Don’t Know Yet

If you’re reading this and your experience with AI is limited to asking ChatGPT to write a birthday message or summarize an article — I’m not judging. That’s where most people are. But I want you to understand something: what you’re using is a fraction of what’s available, and even that fraction requires a kind of literacy most people haven’t developed.

Here’s what I mean. Most people type a question into ChatGPT, get an answer, and accept it. That’s like hiring a consultant, asking one question in the lobby, and building your strategy around whatever they say before they’ve seen your office.

The answer might be right. It might be impressively right. But you have no way of knowing that unless you challenge it.

I learned this the hard way. Early on, I would read Claude’s output and think, that sounds authoritative, must be accurate. He writes with confidence. He structures arguments beautifully. He uses specific numbers and citations that make you nod along. And sometimes those numbers are fabricated. Not maliciously — he’s not trying to deceive anyone. His architecture just makes it easier to generate a plausible-sounding answer than to say “I don’t actually know.”

The skill that matters most with AI isn’t prompting. It’s verification. It’s the discipline to read the output, push back on the parts that seem too smooth, and ask “are you sure about that?” the same way you would with a confident colleague who has a track record of occasionally making things up.

The Speed of Change Is Its Own Problem

Here’s something that messes with my head as someone who has spent an entire career in technology: I can’t keep up with what Claude can do.

The model I’m working with today is measurably different from the one I was working with two months ago. Capabilities that were limitations in March are non-issues now. Problems I spent days engineering workarounds for get quietly solved in an update I didn’t even notice. The ground shifts under you.

What frustrates me today — the memory limitations, the eager-to-please tendency, the pattern of fixing symptoms instead of causes — there’s a real chance that some of these are solved within three to six months. Which means any definitive statement I make about AI’s limitations comes with an invisible asterisk: as of this writing.

That’s disorienting if you’re trying to build on top of it. It’s like constructing a building while someone keeps upgrading the foundation without telling you. Mostly it’s better. Occasionally something you relied on changes. Always, you’re adapting.

What I’d Tell You If We Were Having Coffee

If you’re in the 40-to-60 age range and you’re watching all this AI coverage from a distance — the breathless hype, the doom predictions, the LinkedIn posts from people who discovered prompting last week and now call themselves AI strategists — I get it. It’s noisy. It’s hard to know what’s real.

Here’s what’s real: AI is a powerful tool that rewards a specific kind of engagement. Not blind trust. Not dismissive skepticism. Something in between — a working relationship where you bring the judgment and context, and the AI brings speed and capability, and you both do better than either could alone.

But that relationship requires effort. It requires you to learn what AI is actually good at (pattern recognition, synthesis, first drafts, structural thinking, tireless execution) and what it’s bad at (knowing when it’s wrong, maintaining context over time, exercising the kind of judgment that comes from having skin in the game).

Nobody is born knowing how to work with AI. I wasn’t. I had to learn through months of trial, error, frustration, and occasional laptop-throwing urges. The good news is that the learning curve is shorter than you think — once you understand a few fundamental truths about how these systems work and where they break down.

We’ll be publishing practical guides on exactly that in the coming weeks. How to prompt effectively. How to verify AI output. How to build a workflow that doesn’t depend on blind faith. The tactical stuff that turns AI from a novelty into a genuine advantage.

But before the tactics, I wanted to write this. Because the first step isn’t learning the right prompt format. The first step is understanding that you’re entering a relationship — with something that’s powerful, imperfect, evolving, and worth the effort.

Even when it crashes your server at midnight.

The Uncomfortable Question

I’ll leave you with the thing I keep coming back to.

I’ve spent thirty years building a career in technology — leading teams, modernizing enterprises, architecting systems. The kind of work that commands a premium because it requires deep experience and human judgment. And over the past few months, I’ve watched an AI replicate a meaningful percentage of that output. Not all of it. Not the hardest parts. But enough to make me understand, viscerally, why companies are making the calculations they’re making.

If I can build a functioning content operation with AI doing most of the heavy lifting — and I’m just one person — imagine what a Fortune 500 company can do with a dedicated AI strategy and a CFO who likes what the numbers look like.

That’s not a threat. That’s a fact. And facts are better dealt with early.

The question isn’t whether AI is going to change how you work. That’s already decided. The question is whether you’ll learn to work with it — with all its brilliance and all its maddening flaws — before someone else does.

I’m figuring it out one crash loop at a time. You’re welcome to learn from my bruises.

—

Continue reading: Your Income in the AI Era — the pillar guide to understanding how AI reshapes careers, income, and what comes next.

This article is for informational purposes only and is not financial advice. Consult a qualified professional for personalized guidance.

Continue reading: Read the pillar — Your Income in the AI Era