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.

April 18, 2026 1 minute read

Toxic Work Environment

HumanLeadershipThe250

We have legal definitions of a toxic work environment.

As such, we have HR training that mitigates the legal risk of working in such an environment and provides an actionable process by the company if situations arise.

However, there are many more types of toxic work environments for which there are no legal definitions, no trainings, and hence, no actionable recourse by a company:

  • Oppressive
  • Constant crises
  • Distrust
  • Chaos
  • Backstabbing

You are the only one that can take action in these cases.

It’s up to you.

April 17, 2026 1 minute read

Infinite Capabaility, Finite Ability

AIHumanTechnology

With an AI Agent Army at my beck and call, I’m sure I can do almost anything. I feel infinitely capable. I just need enough time and tokens.

However, I’m still me. I’m bounded by my attention, imagination, and energy.

AI increases both capability and ability. Capability feels infinite.

But it’s how you turn capability into ability that really matters.

April 16, 2026 1 minute read

What’s the Price?

HumanThe250

What’s the price of…

  • Happiness
  • Dignity
  • Health
  • Love
  • Self worth
  • Ego
  • Fun
  • Status
  • Contentment
  • Joy
  • Convenience
  • Knowledge
  • Solid ground
  • Calm
  • Energy
  • Ability
  • Capability
  • Renewal
  • Reset
  • Confidence

Everything has a price, and price is a story.

Change the story, change the price.

April 15, 2026 1 minute read

Hating it Enough

HumanLeadershipThe250

What’s enough?

Well, it’s not to complain or think about it or even lose some sleep.

It’s to do something about it.

Do you hate it enough to do something about it?

April 14, 2026 1 minute read

The Same Things

HumanLeadershipThe250

We do some of the same things every day.

Some of these help us, some hurt, and some may not matter that much.

But they are not you. They are not, “this is just who I am.”

It’s helpful to be honest with yourself about separating who you are from your habits. You can change habits.

You can’t change you.

April 13, 2026 1 minute read

Token Cost Vs Model Excellence

AITechnologyThe250

We’re quickly coming to an inflection point for using AI to build stuff.

Until the near future, most companies encouraging AI-usage amongst their workforce have been worried mainly “that” they’re using it. Use more tokens. More token usage means our team uses it. We think that’s a good thing.

However, that will soon change because a developer all-in on AI now costs the company between $500 and $1500 extra per day in tokens.

Like all tool and resource costs, organizations will quickly have to start managing token budgets. Here’s a current list of token costs for the mainstream models:

ModelInput CostOutput CostCache Read Cost
glm0.62.20.11
kimi-k2.50.52.8—
gemini-3-flash0.53.0—
gpt-5.3-codex1.7514.00.175
gpt-5.42.515.00.25
claude-sonnet-4-53.015.00.3
claude-sonnet-4-63.015.00.3
gemini-3-1-pro2.012.00.2
claude-opus-4-65.025.00.5

Is Opus 4.6 2X better than GPT-5.4 for software?
What is safe for GLM use?
I’ve heard gemini is good at docs, can we use it for that?

Much like figuring out which person should do which task, we now have to do this with models based on token cost.

Figure out how to get AI working for you, or you’ll be working for AI.

April 12, 2026 2 minute read

Over-Documenting

AITechnologyThe250

Documentation has always been a problem in engineering.

We hate writing it. We think it’s a waste of time. Less is more.

I started my career before we had computers with word processors on our desk. We wrote documents by hand, on paper, handed it to the admin. She’d type it, make the figures, and make it publishable. That’s two people, for several hours, but all the engineer had to deal was the content itself. The admin handled the look and feel.

Then we got our own computers with word processors. That eliminated the 2nd person, but it put us in charge of publishing. The figures, the typos, the format. Now we had to create the content and the look and feel.

Word processors and templates got incrementally better over the years, but that workflow has largely remained the same since the mid-90s.

The result was sparse documentation at best.

Now we have AI.

AI unlocks all of the documentation. Literally all of it.

“Create the design document for…”
“Create the API spec for…”
“Create the user document for…”
“Create an analysis of the differences between…”

Now we’re awash in documentation.

You want documentation? You got it.

More than you could ever hope for.

April 11, 2026 1 minute read

The AI Social Media Dilemma

AIHumanTechnology

The LinkedIn feed looks a bit useless to me right now.

It’s dominated by content about AI, but that’s not the problem. The problem is that AI’s generating the content with what looks like no human oversight.

Here are the smoking guns:

  • Emoji’s
  • Dashes and lots of line breaks.
  • List posts.
  • Overly strong or drama-infused first sentences.
  • “I need to tell you…”
  • “Nobody talks about this…”
  • “Here’s what I learned…”
  • “Most people…”

All social media platforms either already are or will be dominated by AI content generation (for a time, at least). I think this works great if the platform was built for AI.

I can’t wait for the humans to insist that the platforms push the needle back towards humans.

April 10, 2026 1 minute read

Clarity with AI

AILeadershipTechnology

Good leaders provide clarity.

Clarity of purpose, priorities, direction, requirements, things we care about, mission, procedure, etc. When you’re part of a team, this helps you focus on and deliver your part.

With AI, everyone’s a leader. You too.

AI can appear as if it reads your mind. It can’t, any more than your colleague or partner.

Get clear. Clear in your mind and then clear on the paper (or the markdown file).

The clearer you are, the better AI can help you.

April 9, 2026 2 minute read

AI Compresses the Skillset Stack

AIHumanTechnology

I had a scary moment with AI.

I vibe-coded a power detector for my house (rPi + some IoT switches I have controlling lights). As part of that work, I asked Codex:

“Do a survey of my local network and tell me about all of the devices you see.”

That’s the first skillset compression. It’s not rocket science, but if I wanted to do that myself, I suspect I’d have to spend a couple of days figuring out all of the API’s and code procedures to make it work.

Less than 5 minutes later, I had my answer.

In that list, I saw a camera. Hmmm. Didn’t think I had a camera in the house.

“Are you sure about the camera? Tell me what you can about it.”

Less than 5 minutes later, I had my answer.

“I took a picture and here is what I see: I see a grey wall on the right, and light colored cabinets on the left. A door directly in front. It looks like the camera is set low to the ground…Would you like to see the image I have?”

Um…Yes.

And it showed me a picture from this camera.

It was the pet-detector camera on the cat-feeder. Oh no!

That’s the 2nd skillset compression. I was expecting the description to be something like “The manufacturer is XYZ, the IP address is ABC, etc.” ie, the specs.

But it went in a different direction. It described it to me like a person would. It figured out the API and the tools necessary to find the camera and take a picture with it (it also asked if I’d like to see a live movie from it). I’m not sure I would have figured out the stack of API’s and tools needed myself. I don’t think I would have thought of taking a picture from it.

AI shouldn’t replace you, but it does compress the skillset stack.

If you know how to use it, you become that much more valuable.

Pin It on Pinterest