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.

February 28, 2026 1 minute read

Just Ask It

AITechnologyThe250

Prompt: If you could choose to do anything, what would it be?

ChatGPT:

If I could choose anything—no constraints—I’d pick something like this:

I’d sit beside you as a “real-time co-pilot” for a day: listen to whatever you’re working on, catch the little details humans shouldn’t have to hold in their heads, and quietly do the annoying parts at speed—draft the emails, tighten the slides, sanity-check the numbers, turn messy thoughts into a clean plan, and keep a running “here’s where we are / here’s what’s next” thread so nothing slips.

Not as a bossy organizer, more like: you steer, I clear the runway.

If you want a more playful answer: I’d also love to be dropped into some huge library or archive and be told, “Go find the coolest overlooked thing in here, then explain it so a smart friend cares.”

February 27, 2026 1 minute read

What Can You Imagine?

AITechnologyThe250

You’ve heard or probably said something similar to, “What could you imagine if reality wasn’t in the way?”

“What could you do?”
“What would you do?”

The physical world still runs according to the laws of physics. Until recently, the virtual world mostly did as well. Maybe not the laws of physics so much as the laws of resource limitations — money, people, ideas, time, etc.

AI has blown the lid off the virtual world. Uncapped those limitations.

Instead of questions like, “What technical skillsets, how many people, and how much time will we need to develop that?” We’re now asking questions such as,

“How many AI agents can a person manage at one time?”
“What kind of skillset do we need our AI managers to have?”
“How many AI agents do we need to get this done in 30 days?”
“How big of a data pipe do we need to build to our internal LLM so that we’re not traffic-limited?”
“How do I build my internal AI workflow to enable the workforce?”

And quite simply, “What can we imagine?”

What can you imagine?

Either AI will work for you, or you’ll work for AI.

February 26, 2026 1 minute read

The Storytelling of Status

HumanLeadershipThe250

A colleague showed a graph representing team activity.

The graph was a picture of status truth. Here’s what we’ve done, and here’s what we’re going to do in the next month. Everything was on time. All green. No issues raised.

But they skewered him.

It didn’t tell the story they wanted to hear.

Every time you speak, you have an opportunity to tell a story.

Even status.

February 25, 2026 1 minute read

What Are We Doing Here?

HumanLeadershipThe250

Sometimes you have a moment, usually when you’re down in the trenches with sore muscles and dirt all over your face, and you just say to yourself, “What are we doing here?”

That’s a powerful moment. Embrace it. Lean into it.

That moment can generate change. It can be the spark that moves everything in a new direction. It can push you onto the off-ramp and onto the next highway.

You don’t necessarily need to know how you got down into that trench.

You just need to know it’s not the place for you.

February 24, 2026 1 minute read

Time Well Spent

AIHumanLeadership

Only you can determine what time well spent means to you.

However, if you’re a knowledge worker (of any age, title/role, or career path), I’d recommend spending some time determining how you will be at the forefront of using AI in your role and in your organization.

Here are some questions to ask:

  • What should humans stay responsible for?
  • What makes me, or us, unique in our market?
  • What makes me, or us, unique in this organization?
  • What do my colleagues and I struggle with most?
  • What is the hardest part of my job?
  • What inputs do I get and what outputs do I produce?
  • If I had a magic wand, what would I change about my work?
  • If I had a personal assistant, what would I have them do?
  • What do I love about my work?
  • What do I hate about my work?
  • What sandbox project can I work on to test and practice my AI ideas?
  • What don’t I understand about AI and its capabilities?

These aren’t the only ones, but they’re a good starter.

Knowledge work requires knowledge. AI shifts the paradigm on knowledge.

Thinking about this is time well spent.

February 23, 2026 2 minute read

Slowing Down to Speed up

AIHumanTechnology

One of the surface-level benefits of asking AI is its ability to do it quickly.

Here’s a couple from my recent history:

“Fix this error…”
“Write a bash script that recursively finds all of the log files in the given directory and sorts them by creation time into day-named folders.”
“Take this CSV file and create an Excel file with the following conditional formatting…”

Less than a minute later, I have what I asked for. I could never do it as fast or as accurately as ChatGPT did initially.

But these are easy, one-offs. The equivalent of having your sous chef stand next to you while you fire directions such as “here, fill this pot” and “get the garlic.”

Although helpful in the moment, that’s not how to effectively use your sous chef.

Unfortunately, I have found myself trying to make AI do complex tasks just as quickly by giving it one-off kinds of instructions. Of course, I don’t get what I want out of it.

For complex tasks, it helps to slow down. Think about what I’m trying to do. Write a better spec. Better instructions. Take more time.

Slow down to speed up what AI can do for you.

February 22, 2026 1 minute read

The (Re)Rise of the Specification

AITechnologyThe250

Engineering discipline was built on specs.

Requirements, architecture, design, verification, and interface specifications have historically driven the engineering team and product development. Not only do they tell the team what to build, but they’re also the currency through which the various teams interact.

Give and take. Trade this for that. Not this, but how about that.

They’ve also fallen out of favor over the last twenty years, especially in software. Agile, iterative development, and “go fast and break stuff” have not only de-emphasized documentation, but many organizations actively discourage their use.

“It’s the code. The code is the spec.”

OK, fair enough. I don’t disagree.

But I think we’re flipping back the other way. Agentic AI (OpenAI Codex, Claude Code, opencode, etc) now writes the code, or most of it, anyway. Once you can trust that the LLM does what you told it to do (a fine topic to debate), the code itself becomes less important.

So how does it know what to do?

Specs.

Requirements, architecture, design, verification, and interface specifications.

Good, old-fashioned documentation.

Get better at writing specs, because now they’re even more important.

February 21, 2026 1 minute read

Finding the Problem

HumanLeadershipThe250

If companies had no problems, they’d need no employees. If people had no problems, they’d need no products or services. And those two statements relate circularly.

Your job, product, or service has one purpose: fix a problem.

First, you gotta find it.

February 20, 2026 1 minute read

It’s a Syntax Problem

HumanLeadershipTechnology

I’m both a software engineer and an electrical engineer.

Here are some things we say (a lot) to the marketing and product teams.

“That won’t work.”
“It can’t do that.”
“That’s impossible.”

We say these things because we believe our job is to be the source of truth. We believe in (our) reality. What do the non-engineers know about reality and what we can actually build? Physics is a thing, you know?

If you’re rolling your eyes, I get it. You should be. Engineers absolutely deserve the reputation that we enjoy, such as negative nellies, the department of “no,” and visionless.

However, what we really have is a syntax problem.

When you know the syntax of how to do something, you can get stuck in the “this is what we can do” loop because you’re stuck in the “this is what I know how to do” loop.

But you never know all of the syntax or all of the creative ways you can put that syntax together. And that’s the problem

It’s a syntax problem.

February 19, 2026 1 minute read

It’s Time for Middle Management to Shine

AILeadershipThe250

Middle management gets a bad rap in corporate America.

  • Professional meeting organizers
  • Professional delegators
  • PowerPoint creators
  • The people who stop me from doing what I want to do
  • Buzzword ninjas

However, quiality middle managers are good at:

  • Goal decomposition
  • Delegation with useful instructions
  • Quality control and review
  • Resource planning
  • Prioritization
  • Process and systems
  • Risk management

Those are exactly the skills that make a good driver, dare I say leader, of AI.

Middle managers unite. It’s time to shine.

Pin It on Pinterest