This series has had its heavy parts. The bottom rung of the career ladder is under real pressure. Teams are getting smaller. The craft a lot of us built our identities around isn't the scarce skill anymore.
All of that is true. So is this: it has never been a better time to be someone who builds things.
The world wants more software, not less
Every wave I've seen in this industry came with a headline about programmers being replaced. In 1981, a product called The Last One was sold as "the last human-produced program that needs to be written." It wasn't. Every wave since has ended with more software and more people building it.
This one is already following the pattern. Since Claude Code launched in early 2025, Indeed says software developer job postings are up almost 15%, while job postings overall fell 7%. Most of that growth is in senior roles, the people who can decide what to build and direct the work. To be fair to the data: postings are still about 27.5% below where they were before the pandemic, and Indeed is careful to say the timing doesn't prove AI caused the rebound. But it's the opposite of the collapse the headlines predicted.
When software gets cheaper, people don't want less of it. They want more. The demand is for engineers who can direct the machines.
Building has never been cheaper to try
Here's the part that still amazes me. The cost of trying an idea, of getting from a plan to working, tested code, has collapsed.
The frontier labs need users, and they're pricing like it. For a sense of scale, OpenAI says its median researcher now uses more than $600 a day of inference at API prices. A $200-a-month subscription isn't the same thing. It has usage limits, and it buys far less compute. But it buys more engineering capacity than I could have hired with a real budget ten years ago, for the price of a nice dinner.
To be clear about what got cheap: producing code. Hosting, maintenance, security and your own time still cost what they cost. But the price of an experiment, of finding out whether an idea works, has dropped to almost nothing. It won't stay this cheap forever. Right now, it's a steal.
What I build
I've fully embraced this. Every project I take on now has an agentic side to it. Two of them show how far that goes, and how fast it's moved.
Retro Golf: the test
I've never been a game developer, but I always wanted to be one. In December 2025, with an early draft of my workflow, I decided to put it to the test: could agents build a game from start to finish, without me understanding or even knowing the code?
They could. Retro Golf is a pixel-art golf game, built in TypeScript with the Phaser game framework and wrapped as an iOS app. In about two weeks it had a swing meter, slope physics for putting, rocks the ball bounces off, water hazards, multiple courses and a driving range. It's live on the App Store, and as of October 2026 it averages more than 50 downloads a day, with all five-star reviews. Publishing was new to me too. The App Store build and submission process is its own maze, and the agents walked me through it until the game was live. The cost was the subscription I already had, plus Apple's developer account.
That game is what made me believe I could build anything.
GameDay Scoreboard: the payoff
Ten months later, the workflow had matured, and I pointed it at a problem of my own: texting volleyball scores to my family during games. I built GameDay Scoreboard in two days, without writing or reviewing a line of its code, at no cost beyond the subscription I already had.
At its first real game, I shared the live scoreboard with more than 15 people. Everyone said it just worked. I posted about it on X for the same reason I wrote this series, to push back on all the doom and gloom: "I challenge every developer who feels the doom and gloom of AI to embrace it. Look at the dumb problems and toil in your life, then write software to solve them."
What happened next
Honestly, nothing dramatic. Neither project had a breaking failure. What they had was a steady stream of small fixes and quality-of-life improvements, and each one made the agents a little better at understanding my work.
That's how I run everything now. I start with agents, and I improve their understanding of my work every day. It's how I keep track of everything I work on, not just the apps I build.
And the rest
- This blog, and Jordan Pushed Off, my Utah Jazz fan blog. Both built, designed and deployed with agents.
- My company's books. Not everything is an app. My agents keep a complete double-entry ledger for my company using Beancount, a plain-text accounting system I knew nothing about before. The agents did. Every transaction is auditable, I can review and reclassify anything month by month, and my accountant reviews it each year. That used to be days of my life.
None of these started with "I want to learn agentic engineering." They started with something I wanted, or something annoying, and a decision to stop putting it off.
Why your own problems are the best place to start
Building something for yourself is the fastest way to learn everything in this series.
You write the plan, so you learn to decide what's worth building. You approve the requirements, so you learn to spot the gaps before they're expensive. You own the outcome completely, because you're the user. If the scoreboard breaks during a game, I'm the one in the bleachers who notices.
It's also the safest place to learn. The stakes are low, and because you're the user, you find out fast when something's wrong. The lessons transfer straight to work.
The challenge
So here it is.
Look at the small, dumb problems in your life. The thing you do by hand every week. The spreadsheet you keep meaning to fix. The update you keep texting to the same five people. Pick one, and build the software that makes it go away.
Then do it again. Build something every month, or every week, or every day if you can. Run it through the process every time: plan, requirements, implementation, docs. Ship it. Use it. Fix what slips through, and fix the process that let it.
Set yourself a budget, and spend it on building things instead of worrying about them. And who knows, maybe one of these projects takes off and turns into real side money. Until then, I can tell you from experience: building and shipping things is the most fun this job has ever been.
Now what?
I started this series with a question: if coding is solved, now what?
Here's my answer. Now you build. Not by typing every line, but by deciding what's worth making, setting the bar, owning the result, and putting the machines to work on the rest.
The code was never the point. Solving problems was. And we've never been able to solve more of them.