🏃 Jane, Stop This Crazy Thing: The Buy-Make-Buy Treadmill

"Jane! Stop this crazy thing!"
— George Jetson, every episode, usually while strapped to a malfunctioning treadmill

The Jetsons promised us a future where the robots do the work and humans push one button for a few hours a day. Spoiler: George still got yelled at by Mr. Spacely, and the treadmill still ran away with him.

I envision myself on that treadmill every time I sit in a roadmap review. The robots showed up, the button-pushing showed up, and the part we got exactly right was the guy flailing on the belt while someone else controls the speed. (Wave if you've been George this quarter 👋)

The Cycle, Now at 4x Speed 🔁

Here's the loop every one of us has lived through:

1. Open source or self-built, because it's free
2. Paid service, because free turned out to cost us time
3. Back to open source or self-built, because the bill got ugly
4. Back to paid, because the pager got uglier

Repeat until someone gets promoted for "modernizing."

It used to take years. You'd adopt, regret, migrate, regret differently, and by the time you finished the loop, half the org that started it had moved on. Nobody noticed the circle because the circle was slow and no one was here to remember.

AI did the loop in twelve months. Open-source models, then paid APIs, then in-house builds, and we're about two quarters from the CFO asking why we're running GPUs when a vendor will do it for a line item. Same treadmill. Faster belt.

That's the useful thing about AI as a case study. It's tech adoption in fast-forward, so you can finally see the shape of the thing.

My Best Friend's Argument (And Why It's Half Right) 🤝

This came up twice in one weekend, in two separate conversations, and both landed in the same place. My best friend made the fair counterpoint: this is just the market working. Price and performance swing. Open source catches up, then falls behind. Vendors get cheap, then get greedy. Responding to that is called engineering.

He's right. Costs and capabilities really do move. Teams really do respond to forces outside their control.

In engineering terms: the trigger is real. The response is the problem.

Because here's what I see in the actual rooms:

• 🚀 Switching before the pain, because the demo looked good
• 🧪 Committing before the experiment, because the keynote was loud
• 👑 Chasing the frontier at a company that has no business being at the frontier
• 🪞 Blaming the tool for a problem that lives in the team

Responding to a change in the market is engineering. Responding to a change in the conference schedule is cosplay.

Somebody's Dream of Being the Pioneer 🧭

Every migration has a sponsor. And a lot of sponsors aren't fixing a problem. They're auditioning for a role.

The pioneer. The one who saw it first. The savior who moved the company to the future before the competition did. (Somewhere, a Spacely Sprockets exec is sweating over what Cogswell Cogs just announced.)

Pioneers are great. Pioneers also died of dysentery in large numbers. The Oregon Trail was not a fun game for the people in it, and we still pick "ford the river" every time.

In other words:

Most companies aren't the tip of the spear. They're the guy holding the spear's gift receipt.

Your boring infrastructure that half the industry has run successfully for a decade, the one you keep complaining about? It's probably fine. You're probably the problem. Your config. Your ownership. Your on-call hygiene. Your habit of never finishing the last 10%.

Nobody wants that answer. It doesn't have a launch event.

What Actually Changed 😮‍💨

Fear. That's the engine now.

Fear of missing out.
Fear of being left behind.
Fear of being found lacking.
Fear of being the imposter in the all-hands who didn't have an AI strategy slide.

Fifteen years ago, I'd have told you this industry ran on something else. Idealism. Delivering real value to real people. Arguing about whether a thing was right, not just whether it was shippable. People who'd stand up in a design review and say "we shouldn't build this," and be heard.

That's not nostalgia talking. Okay, it's a little nostalgia. But I remember the arguments, and they were better arguments.

Now we've flattened the craft into a single word: coding. And once the job is "coding," the headline writes itself: the robots code, so the role is obsolete. Neat. Tidy. Wrong.

Coding is the typing. Engineering is knowing what not to type. It's judgment, tradeoffs, scar tissue. It's the thing that tells you the shiny migration is a six-month detour with a hidden tax. A robot can write the function. It can't (yet) tell your VP "no, and here's why" in a room where nodding is the safe move.

🪦 Sunset Policy → Company Edition

Here's the line I keep hearing: software engineering roles are evolving toward obsolescence, and that's just the natural lifecycle of an industry. Tools change, the job changes, some jobs go away. Adapt or age out. Said calmly, from a stage, by people who would never say it about their own seat.

Fine. Let's apply the standard evenly.

Another natural lifecycle is the fall of a company. Birth, growth, plateau, decline. Acquired, hollowed out, wound down, or just quietly outcompeted by someone who showed up hungrier. Companies die. They always have.

In engineering terms: every service we ship gets a deprecation path. We write the sunset notice. We migrate the users. We turn off the lights on purpose, because the alternative is a zombie service nobody owns that pages somebody at 3am. The company is the one system in the building with no decommission plan.

And that's where the fear comes from. A company that can't accept it will end makes frightened decisions. The AI pivot becomes a defibrillator. "Reimagining the business" becomes a way to avoid saying the core product is mature. Mature isn't dying. Mature is paid for. But fear reads maturity as the first symptom, so the treadmill gets another notch.

What it looks like when a company makes peace with it:

• 🧾 Be excellent at the thing you're actually good at, and stop apologizing for it
• 🌅 Know what a good ending looks like: a profitable harvest, a clean merger, a wind-down that treats people right
• 🧪 Fund experiments from surplus, not from panic
• 🤝 Hold your people to no standard you wouldn't hold the company to

And here's the ironic part. The company that accepts its own mortality is the one free to make smart decisions. No need to be the pioneer. No need to win the arms race. No need to bet the runway on a hype sprocket to prove it's still alive.

You can't ask your engineers to make peace with obsolescence while the company burns its runway to avoid its own.

The Jira Board Is Not a Chore Chart 📋

Small side rant, but it's the same disease powered by fear.

Somewhere along the way we started putting housekeeping on the board. Upgrade this. Clean that up. Rename the thing. Tech debt tickets that are neither a bug nor customer value, just maintenance, now estimated, groomed, prioritized, and demoed.

In engineering terms: we turned hygiene into a process.

You don't schedule brushing your teeth in a sprint. You don't story-point doing the dishes. Maintenance is the job. The moment it needs a ticket and a stakeholder, it's not a chore anymore. It's a negotiation. And we are terrible at negotiating on behalf of the boring stuff.

The Pretzel Ring Problem 🥨

I have two kids. Here is a real conversation I've heard, more than once, in my own kitchen:

"I have more pretzel rings than you."
"I finished my milk first."
"That's not fair, you started first."

I have sat in meetings with that exact energy. Different vocabulary. Same stakes. We moved to the new model before the other team did. We've got more agents in production. We shipped our own, so we're ahead.

Ahead of what? Nobody says. The finish line isn't on the map. The only thing everyone agrees on is that stopping means losing.

I've been doing this for more than fifteen years. I started as a graphic designer, taught myself my way into code, wrote Objective-C when that was a personality, shipped an iOS game on Cocos2D, lived through the Rails years, and rewrote things I'd sworn were final. I've been on both sides of the buy/build argument, and I was sure both times.

I was right about half of them. The half I got right, I'd waited. The half I got wrong, I'd been excited.

And I hate how that sounds when I say it back. I don't want to think the people making these calls are making dumb calls. I don't want to look at a leadership decision and see my kid and his snack bag. I want to believe we have smart people making smart decisions for good reasons, like I used to.

But I keep seeing the same thing from everyone I talk to, at different companies, at different sizes. My best friend says maybe I should move on. I'd love for this to be a company problem. It isn't. It's a climate.

Who's Actually Winning the Arms Race? 🏁

Follow the money, because it's the most honest part of the diagram.

Every team racing to adopt, switch, rebuild, and re-adopt is a customer. The loop is not a bug in the market. For a handful of companies, the loop is the product. Keep everyone running, keep everyone anxious, keep the next announcement landing right before the last one gets paid off.

A company on a track, running in circles, to the applause of the few people selling the shoes.

So why are we racing? What's the prize? Nobody asked for one. We just heard the starting gun and ran.

Boring Is a Strategy and an Art

Here's the rule I'd like to see us try. It's not new. It's the one we keep forgetting.

Follow convention.
Then help push convention.
Then adopt the new convention.

Experiment freely. Spike it. Prototype it. Run it on the side project nobody cares about. But commit only when you need to, or when you can point at an objective gain and not a vibe.

Not because new tech is bad. Because the cycle only breaks when someone is willing to sit it out and be fine.

Be fine with being mature.
Be fine with being boring.
Be fine with being finite.

We didn't lose the craft. We just stopped defending it.

The treadmill only stops if someone's willing to step off.