johnmaconline chain logo johnmaconline / The250
Daily one-pagers about AI, leadership, technology, and being human.

johnmaconline

I'm writing to think, learn, and remember in public. I'll be here everyday.

November 5, 2026 2 minute read

Whales in a School of Fish

AITechnologyThe250

“In the 20th century, we were trained to be whales—massive, slow-moving, centralized, dominant. In the 21st, power has shifted to the school of fish: agile, decentralized, coordinated in real time.”
– Lisa Gansky

At this point, it’s a cliche to say “AI changes the game.”

It’s true, of course. I’ve been working through that here and taking you along in my daily existence as I wrestle with AI each day. It’s energizing. It’s exhausting. It’s fun. It’s scary. But make no mistake, you better figure out how to get AI working for you, or you will be working for AI.

As soon as I hooked into ChatGPT around Thanksgiving of 2022, I recognized it as the game changer. I didn’t yet know exactly what it would look like, but I knew. I first started talking about using it versus it using you here.

Which brings us to fish and whales.

You’re a fish. Institutions, big companies, and governments are the whales.

Consider Microsoft.

Microsoft already has/had the equivalent of 10s of thousands of software engineers that they could throw at any problem or project. Nothing really changes for them.

The magic of generative AI is for you. The fish. It allows you, the indivual, or a small team, to do amazing things. It gives you the ability, which in some ways, gives you the power. It gives you access to an elastic army of superworkers.

What can you do with an elastic army of superworkers?

AI rewards people with ideas.

November 4, 2026 1 minute read

Frivolous Tokens

AIHumanTechnology

I wonder how many tokens were spent this year on some form of this prompt:

“Help me decide what to be for Halloween.”

That, of course, is not the only frivolous way to spend tokens. I’m not sure where the line is. What’s a good use of tokens? What’s a frivolous use of tokens? What can we spend tokens on? What should we spend tokens on?

In fact, is using AI to help you decide what to be for Halloween frivolous? Because who am I to decide?

Just because you can doesn’t mean you should, but sometimes you should.

November 3, 2026 1 minute read

Don’t Miss Them

AIHumanTechnology

The little moments.

Life distracts us. Social media. A million AI agents running. The News. The burdens. The existential dread.

But each day, if you’re tuned in, life presents little moments that counteract. That help you feel human. That help you feel.

Tune in. Don’t miss them.

November 2, 2026 1 minute read

Herding Agents

AITechnologyThe250

With all of the ideas flowing, agents working, and terminals spewing, it’s difficult to keep track.

Asking each what’s going on is one approach. Quite frankly, it’s annoying and not very efficient. You/me quickly become the bottleneck.

I need something that can look at it all and show me what’s happening where.

It’s time to get serious about herding agents.

November 1, 2026 2 minute read

The Joy of Building

AIHumanTechnology

There’s something deeply satisfying about building something that didn’t exist before.

Software engineers know this feeling. So do artists, carpenters, metal fabricators, gardeners, and anyone else who makes things.

That satisfaction comes from lots of places.

Having an idea and making it real. Getting lost in the flow of the work. Being an expert at something. Solving a problem. Finding that elusive bug. Discovering an elegant solution. Seeing something work for the first time.

And, of course, looking at the finished product and thinking, I made that.

From the dawn of computers until about the end of 2025, software engineers got to experience all of these things. And forming the syntax in your brain, then typing it in, wasn’t just a necessary part of the work. It was part of the reward.

That part has died. Forming the syntax and typing it in. You don’t do it anymore. Let’s put its time of death at January 2026.

AI does that work now.

It also does a lot of the debugging. And the designing. And the testing. And the refactoring. Often faster and better than we can.

For many engineers, this feels like losing the best part of the job. I was in a meeting yesterday where someone said, “But that’s what I love about my work. How will I derive any satisfaction from what I do?”

I understand that point of view. But I don’t share it.

Was typing code really the best part? The syntax?

Or was it the satisfaction of solving the problem? The creativity of designing the solution? The judgment required to make the tradeoffs? The curiosity that led to understanding something new? The actual functioning thing?

We have no choice now but to separate the satisfaction of creating from the mechanics of creation.

The software engineer’s job isn’t disappearing. But the source of satisfaction is moving.

From writing the code to imagining what could exist. From debugging the implementation to questioning the design. From mastering a programming language to mastering a problem domain.

The challenge isn’t just learning how to work with AI.

It’s learning to find the joy of building in the other areas.

October 31, 2026 2 minute read

Both Ends of the AI Candle

AIHumanTechnology

A standard candle has two ends, of course.

One end is for burning, and the other connects it to the earth. We have an idiom about using that connection end for burning.

AI gives us two levels at which we can burn both ends of that candle.

The first is us.

For people with ideas, or maybe just under pressure, AI makes it easy to burn both ends. Just get another agent running. Open another tab. More ideas. More instructions. More.

It’ll happen if you let it. The franticicity. There’s a reason why we, as humans, need one end of that candle firmly connected to the earth.

The second is AI itself.

Can AI get frantic? Can it work too hard?

I’ve seen it get confused. I’ve seen it cross paths and step on other agents doing work for me. I’ve seen it get into the death spiral. But that’s more of a me problem than an it problem. Poor instructions and expectations. That looks a lot like a standard management problem. People and agents both respond to better management.

There’s also the token budget. AI runs out of tokens. Burning both ends means more tokens, which means budget overflow faster. That’s generically a money problem. More money fixes it.

Then there are the big picture concerns. Power. Water. Noise. Real estate. The more we have AI burn both ends, the more we need and get of each of these. These are the “should we?” questions that rightfully are front and center in our culture right now.

I think AI can get frantic and work too hard also, but maybe not in the same way we can.

I’m starting to think burning both ends eventually burns us, even with AI.

October 30, 2026 1 minute read

It Helps if You Know Stuff

AILeadershipThe250

In most professional settings, truth and facts matter.

It’s OK to say “I don’t know” and follow it up with “I’ll find out” or “I’ll verify with Jim” or some phrase that indicates you’ll get the answer. Those are the right things to say if you’re on the hook, or you’re putting yourself on the hook. You want to be on the hook.

But it helps if you already know the answer.

Professionals know stuff.

AI doesn’t change that.

October 29, 2026 1 minute read

Be Careful With Your Customers’ Trust

HumanLeadershipThe250

I got tricked. Sort of.

I had a subscription that I’ve been trying to cancel, and I thought I had. But alas, I just got the email indicating that this month’s purchase is on its way. I immediately sent an email (no phone number or even an AI chatbot). They replied with “sorry, the order is too far along to recall it now. You may send it back with this return code and a 15% restocking fee.”

Alright, you got me. This time.

The Restore Online Shoppers’ Confidence Act (ROSCA) requires businesses selling online through automatic renewal to:

Provide “simple mechanisms” to stop recurring charges.

That last phrase is the actual language. ROSCA doesn’t prescribe a required maximum number of clicks or method of visibility, but like art, you know it when you see it.

I’m sure someone could walk me through exactly how to cancel my subscription and why they meet the intent of this statute.

But c’mon. I’m annoyed.

You got me. This time.

Was it worth it?

October 28, 2026 1 minute read

AI Sucks at Customer Relations

AIHumanTechnology

“Yes, you’re right, my mistake.”
“I apologize. I missed that.”
“That’s a great idea! Much better than what I was thinking.”

I’ve received each of these as feedback from my AI sessions in the last couple of days. The AI being sycophantic, or at least, trying its hand at the human art of telling me what I want to hear.

It’s bullshit.

And bullshit isn’t good customer relations.

Good customer relations starts with making the customer feel heard. AI is a bit more like, “You’re right. Whatever you just said, you’re right.”

Which makes me wonder whether it heard me at all.

Sometimes I’m wrong. Sometimes my idea isn’t great. Sometimes I’ve misunderstood or missed something. If every correction I offer gets an apology and every suggestion gets applause, I have to do the work of figuring out whether either is actually correct. And most times, when working with AI, correct is what I’m after.

Customer relations is worth being human over.

October 27, 2026 1 minute read

The Code is Not the Product

AITechnologyThe250

It never really was.

Although we treated it as such. The code. The code is the thing. The code is the most important.

But what is the code?

It’s the means to the end. It’s the material. It’s the scaffolding on which a product is built.

It’s the wood in a house frame. It’s metal in your car. The quality is important, of course. Bad wood or bad metal isn’t gonna hold up. It won’t make a good house or car.

The code matters, but it’s not the thing.

AI has put the code in its place.

Pin It on Pinterest