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.

October 4, 2026 2 minute read

Managing

AIHumanLeadership

What makes a good manager?

Is it people skills? Getting the team to deliver on time? Transparency? Communication? Vision?

The only correct answer is, “That depends.”

Is it from the employee standpoint (managing down)? Executive standpoint (managing up)? Shareholder’s standpoint? What does each care about in this specific context?

In many real ways, managing a team of AI agents is very much like managing a team of people. Successfully managing those agents takes many of the same skills. Get better at those skills, and you’ll have more success.

However, there is a big difference. It may seem obvious.

AI agents are not people. No matter what the current argument around AGI and anthropomorphism suggests.

People have feelings. They have experiences, desires, dislikes, bad nights of sleep, children, extended families, partners, and lives.

AI agents have none of that.

As such, managing a team of AI holds a lot of appeal for managers burned out on the people side of the job. A manager of AI can focus on the process and output. The technical delivery. The building of stuff by the team. The question of “what did we get done and how good is it?”

I admit it. I find it appealing sometimes.

Now leadership has choice and opportunity. Manage AI or people? Manage AI and people?

What makes a good manager?

October 3, 2026 1 minute read

Token Drinkers

AITechnologyThe250

Having trouble figuring out which mode to use?

The latest ChatGPT subscription presents you with three basic model options:

  • Astra
  • Sol
  • Luna

Luna is a sipper. Think a well-dressed mid-lifer at an evening cocktail party on the rooftop veranda of a condo on Rittenhouse Square, holding a glass of red wine. They’re not that deep. The conversation is surface-level and guarded. Maybe a bit salesy.

Sol’s having a good time. Think relaxed with good friends or colleagues at their regular Friday spot, working on a few of their favorite cocktails. Maybe there are cards on the table. They’re sharing real stories. Solving real problems.

Astra has a problem. Think the smartest in the room with a bottle of Jack on the table. Napkins covered in diagrams and a laptop with a million tabs open. Slurring their words with no regard for how much this night is gonna cost. But they’re in deep, and they won’t crawl out until either the money runs out or the problem is solved.

Who do you need to help you?

October 2, 2026 1 minute read

The Death of the Software Laborer (The New World)

HumanLeadershipTechnology

What will you do if you don’t have to write the function, develop the data structure, or code up the tests?

Like one guy with the shovel learned how to run the excavator and another why they were gonna dig the hole, you’ll need to learn how to direct the software-producing system and what it should build.

Here’s what a software laborer will still do:

  • understand the problem
  • understand customers and users
  • decide what should be built
  • handle ambiguity
  • understand how technologies can be used
  • break problems into pieces and manage the work
  • choose architectures and tradeoffs
  • determine whether the output is actually correct
  • test against reality
  • integrate systems
  • decide when not to build something

Not many construction workers today pine for the time when they used a shovel all day.

Neither will software laborers pine for the time when they typed in code themselves.

October 1, 2026 2 minute read

The Death of the Software Laborer (The Resurrection)

AIHumanTechnology

We generally don’t need thousands of people with shovels anymore.

But that doesn’t mean we don’t need those thousands of people. We just scaled them up and across the stack of work. Now need them to

  • operate the machines
  • survey and lay out the work
  • install pipe, conduit, drainage, rebar, forms, and concrete
  • weld, wire, assemble, and finish
  • maintain and repair equipment
  • inspect quality and safety
  • coordinate crews, materials, and sequencing
  • handle the awkward, irregular work machines are bad at
  • make decisions when reality doesn’t match the plan

The excavator doesn’t decide where it should dig a hole. Nor does it decide what you should put in the hole. Nor why it’s digging a hole. It just digs.

We do need far fewer people per unit of output, but we can produce so much more output. For a job that required 100 people with shovels, today you might need 10 split across the various roles.

And then you can do 10 more jobs at the same time.

AI is the excavator.

We generally don’t need thousands of people typing code into their code editors for a particular project. But that doesn’t mean we don’t need those software engineers. We need our engineers to decide what to build, how it should work, how the pieces fit together, whether it is correct, and what to do when it isn’t.

We don’t need software engineers to write the code.

But boy, do we still need software engineers.

September 30, 2026 2 minute read

The Death of the Software Laborer (The Funeral)

AIHumanTechnology

Each day from 1817 to 1825, across northern New York, 9000 men showed up with their shovels and lunch pails to dig the Erie Canal.

The 363-mile canal crossed New York State from the Hudson River to Lake Erie, spanning 40 feet wide and 4 feet deep with 83 locks. Thousands of laborers with picks, shovels, wheelbarrows, horses, and carts moved the dirt a little bit at a time.

Dug by hand. Men with shovels.

Some were better and faster than others, and that mattered for any specific work crew, but at the project level, a laborer was a laborer.

Need to move more dirt a little faster?

Hire more people with shovels.

But now we have excavators, dump trucks, and tunnel boring machines. What took 8 years could be done in about 18 months today with a few operators and a healthy fuel budget. And those operators bring their lunch pails right into the cabs with them.

Nobody hires a guy with a shovel anymore.

From the beginning of the software industry until today, the same has been true for writing code. Men and women showed up each day with their keyboards and lunch pails to type the code into the machine.

Like the big, strong guy on your work crew, some coders were much better than others, and that mattered for your specific project. But at a large scale, a coder was a coder.

But now we have AI. AI is the excavator, dump truck, and tunnel boring machine.

What took four programmers two weeks to code, test, and fix can now be done before lunch with a few agents and a healthy token budget. Now those coders can eat their lunch while the agents are building the things.

Start the soft music. Put the montage together. Stand the lectern up at the front of the room.

The programmer’s career is in the casket.

September 29, 2026 1 minute read

Robots on the Fasion Runway

AITechnologyThe250

Formless AI is here to stay.

But what about AI-powerd robots?

Sure, we’re worried about Skynet, but Elon’s got Optimus cooking and you can choose from a whole bevy of humanoid personal assistants at the Robostore (the G1 is 17% off right now!). They’re coming.

For some reason we’re building robots that either look robots, or look like women. OK, I guess you don’t have to think too hard about that.

But Vogue pushed the genre forward by sending them down the runway. I guess it didn’t go over well.

The robots are coming, but it’s too soon for the fashion runway.

September 28, 2026 1 minute read

Luddites

AIHumanTechnology

People generally don’t identify as luddites.

We don’t want to be the old guy. We don’t want to be our out-of-touch parents, the crotchety neighbor, or the coworker at the end of his career just trying to ride it out.

Yet, here we are.

AI is an accelerator. It also appears to be accelerating the pool of luddites.

Let’s keep talking. Let’s keep debating. Let’s keep pushing and pulling.

But let’s not stop building.

September 27, 2026 1 minute read

More BS College Ranking Shenanigans

HumanThe250

US News & World Report flipped the script, and it screwed Princeton.

This year’s rankings put MIT at the top. I’m sure the person in charge of this at Princeton is seething because USNWR pushed a methodology change.

This year’s rankings introduced a criterion comparing graduates’ earnings by major four years after college with those who studied the same major nationwide. The measure includes only federal financial aid recipients, part of an effort to minimize the impact preexisting wealth and family connections could have on post-college earnings. It replaces a metric that measured graduates’ federal loan debt.

How dare they.

We should all be very upset.

Oh, right. None of this matters. No rankings list that is fundamentally for-profit, provides its criteria publicly with a wink, and simply asks the institutions to “self-report” on that criteria should matter one iota to you.

Not one neuron operation in your brain should be spent on this list.

Don’t worry, Princeton, you can self-report the proper data next year.

September 26, 2026 1 minute read

Eyes on the Incentives

AIHumanTechnology

Lots of doomsdaying around AI recently. The accelerationist versus doomer wars are heating up.

Dario from Anthropic has always been a doomer, but now his foil Sam, has joined in. Elon has always kinda been a doomer, but then sometimes he just flips to the accelerationist side.

There’s not a clean political split. The left tends towards doomerism. The right tends towards accelerationism. But it’s way more complicated.

What’s really going on?

It’s incentives. Like always.

It’s possible that Dario and Sam are legit in their public worrying. That their incentives are human-focused. But it’s also possible that Anthropic and OpenAI will benefit from a slowdown. From some regulation (that they help create).

Honestly, I don’t know which it is.

But I know I’ll keep my eye on the incentives.

September 25, 2026 2 minute read

Leading AI

AILeadershipTechnology

The current token budgets are untenable.

Or you could say the rate at which the new models use tokens is untenable.

With the standard ChatGPT business account, I can work for about an hour until my 5-hour budget is totally blown. And that’s not even using the frontier model Astra. That’s with the (now old) Terra at medium context. With Astra, its more like 20 minutes.

Work for an hour, then wait four more until my 5-hour budget resets. Then do it again.

So that’s two hours of work. Three if I spread it out across another 5-hour window.

You can’t take off working like this.

What are your options?

  • More tokens. You can always throw money at the problem.
  • Get better at which model to use for what task. You should do this anyway. It helps, but it only takes you so far.
  • Break jobs into smaller tasks. Context costs tokens. Less context, less tokens.
  • Avoid duplication. Models, especially when you have multiple agents running. If you don’t manage what each is doing, they’re probably each rereading all of the same context.
  • Build efficient, AI-readable memory storage. If you cut down what the model needs to read to understand, you use less tokens. This goes hand-in-hand with the one above.
  • Build deterministic software tools to perform much of the work. Use the AI to generate reusable code, rather than asking it to continually use a model to do the same work over and over. This is software development 101. Write code.
  • Do more yourself. Duh.
  • Ask better questions. Give better instructions.

When you look at this list, it looks a lot like what a good manager does with his/her team.

Once again it comes down to being a good leader.

In some ways (not all ways), leading AI isn’t much different from leading people.

Pin It on Pinterest