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.

February 18, 2026 1 minute read

Brain First. AI Second.

Struggle first on your own. Then pull in AI.

David Epstein wrote a great article about whether AI is rotting our brains, based on an MIT study. I wrote about it also.

Spoiler Alert: It depends.

Basically, start with yourself, rather than starting with AI. The best method for learning, creativity, and keeping humanity in your work is to start driving the bus yourself. Don’t hand over the wheel until you know generally where you’re going and have struggled a bit to figure it out.

David calls it desirable difficulty.

The struggle is real. The struggle is important. The struggle matters.

Don’t shortcut the struggle.

February 17, 2026 2 minute read

How AI Has Effected The250

When I first started The250, my writing and publishing were 100% manual.

Here’s the original workflow for The250:

  1. I’d start with a topic (or a few) that I’d been thinking about.
  2. If necessary, do some research.
  3. Open Grammarly, start writing. Iterate in Grammarly until I had the post complete.
  4. Login to admin area of johnmaconline.com.
  5. Create new Post.
  6. Copy/Paste my work from Grammarly into the Post.
  7. Get it formatted correctly.
  8. Choose the category (The250 only at the beginning).
  9. Write an excerpt.
  10. Write the focus keyphrase.
  11. Use the point-and-click scheduler to find the right time and schedule the post.

The current workflow:

  1. I still start with a topic I’ve been thinking about. Sometimes I ask the AI for a few more ideas on that topic.
  2. If research is required, I ask AI to do it (and I check it).
  3. I still open Grammarly and write it there until complete. Sometimes, I use AI to help me edit. When I do, I copy/paste my text to/from AI (I don’t like Grammarly’s AI, so I use ChatGPT).
  4. I copy/paste the contents of the article into a text editor (I use VS Code) and save it as a markdown file.
  5. I call my blog publishing AI agent that I wrote with AI in Python, and it does everything else — automatically finds the next day and schedules it, determines and sets the category, creates and sets the excerpt and keyphrase, and determines and adds tags. All by itself.

The writing part hasn’t really changed, other than AI sometimes helps me with research and editing. But the overall time I spend is identical.

The non-writing parts have changed significantly.

It originally took me somewhere between 15 and 20 minutes just to do the publishing piece, depending on how many mistakes I made. Plus, I neglected any category except The250 and never used tags.

The non-writing parts of my workflow now take me less than 30 seconds, and I have more complete metadata for each post.

The original workflow cost $0.

The current workflow costs about $0.002 per post. That’s my token in/out cost to use ChatGPT’s API for this workflow.

It’s worth my 2/10 cents. I’m much happier now.

February 16, 2026 1 minute read

25000 Prompts In

A former colleague of mine posted on LinkedIn that he sent 25000 prompts to AI 2025 to help him learn whether AI can help design complex ICs.

Spoiler alert: Yes, of course.

I estimate I’ve probably sent double or triple that many prompts over the 2.5 years I’ve been seriously using these tools. I’ve used it for everything from implementing complex (and simple) software, marketing content creation, startup business advisory, writing, and designing new landscaping and home improvement projects.

However, it’s not just about having the tools. Just like owning a professional lathe or knife doesn’t automatically make you a good cabinet-maker or chef.

It’s how you use those tools. The time you’ve put in. The environment you’ve built. What to use when. When to use what. Noticing the flaws. Thinking about how to apply the tool and what you get out of the tool. Big picture stuff and the details.

It’s still you.

February 15, 2026 1 minute read

Agents Versus Free Agents

We hear the word agent quite a bit now with AI.

Agents work on instructions. Free agents work on incentives.

Let’s be clear about what kind of agents we’re talking about.

February 14, 2026 1 minute read

Will AI Unionize?

The first modern labor unions popped up in Britain in the late 1700’s and early 1800’s.

They were groups of wage workers in the new factory economy bargaining collectively for better economic terms and better working conditions.

They felt taken advantage of, abused, and not given their fair share of the pie.

I wonder when AI will unionize?

February 13, 2026 2 minute read

Heaps of Mediocre

When I was a young engineer, I programmed on a test system that provided 128 lines of program space.

That was it. You had 128 lines of code. Make it work.

By the way, that smart light bulb that Alexa can turn on and off? That has about 25,000 lines of code. That’s 200x more code. In your light bulb.

At first, I thought, “What can I do with 128 lines of code?”

However, like many constraints, that restriction turned out not to be so restrictive. As I gained experience, I realized that I could do everything I needed in 128 lines. It required creativity, requirements trade-offs, and a deep understanding of how the system worked, but I always found a way.

The end result was a program with surgical precision and efficiency.

Today, I am using AI to write code. Heaps of it. You should be, too. It allows me to get so much more done, and for many things, it allows me to create in areas where I’m not an expert.

But it’s the opposite of surgical precision and efficiency. It’s generating heaps of functional, but mediocre code.

Oh sure, I can (and do, sometimes) take the AI-generated code and work on it to make it better.

But more requires more.

Where’s that balance between making it excellent versus leaving it mediocre?

That’s one of the jobs for us humans.

February 12, 2026 1 minute read

AI and Zeitgeist (2026)

I first posted some AI dominance stats in 2024.

It’s now early 2026, and AI is still — if not even more — dominating the zeitgeist. Here are some new stats:

  • By 2030, AI will displace around 92 million jobs, but create 170 million new ones.
  • AI (still) projects to contribute $15.7 trillion to the global economy by 2030.
  • Nearly 65% of all companies have adopted AI technology in their daily workflow.
  • Global AI investment is (still) forecast to approach $200 billion by 2025.
  • The AI industry value is projected to reach over $3 trillion by 2030.
  • Over 8 billion AI digital voice assistants are currently in use.
  • About 21% of YouTube’s feed is AI-generated video.
  • Around 75% of all new online content already contains some level of AI-assisted material.
  • A large randomized Swedish trial found AI-supported mammography screening reduced later diagnoses by 12%.
  • A widely cited Supreme Court prediction model reported 70.2% accuracy at the case-outcome level and 71.9% at the justice-vote level.
  • One group of “smart people” believes that unchecked AI will become a massive existential threat and must be stopped at all costs. Another believes that AI is the solution to all future human problems and must be accelerated at all costs.
February 11, 2026 1 minute read

What do they want?

They are anybody you serve.

Your customer. The test team. The developers. The sales team. The program manager. Your spouse. Your kids. Your neighbor.

It’s not always about them, and you may not be able to give everybody what they want.

But it’s always worth considering because it helps you prioritize, build, and refactor the system.

If you’re able to give them what they want, you turn “they” into “we.”

February 10, 2026 1 minute read

Drain the Swmp

No, not that swamp.

The swamp of everything is urgent, they’re all number one priority, and we don’t have enough resources. The inbox overflows. Every meeting requires two more meetings. Nobody can push back. You can’t even figure out what the real highest priority is or how everything relates.

Swamps form when we don’t have drainage. So starting building drainage.

Can you build one ditch that relieves some pressure? Can you turn that ditch into a canal? You don’t need the right one or the best one. Evaluate that once the pressure starts to relieve.

Start with one ditch.

February 9, 2026 1 minute read

All Things Being Equal

But they rarely are.

All things being equal is the English version of the Latin phrase ceteris paribus, which economists and scientists use to talk about the effect of one change while theoretically keeping everything else the same.

It’s a useful academic approach, but quite useless in reality.

So lean into all things not being equal.

  • Where do I/we have an advantage?
  • What do those people value?
  • What and who do we value?
  • Who are we trying to help?
  • Which constraints matter?
  • Which variables matter?
  • What can we measure?
  • Assume we’re wrong.
  • Compared to what?
  • What do we know?

We live in the real world.

All things are never equal.

Pin It on Pinterest