LUMP Depot
We talk a lot about what's happening to junior devs in this world of AI coding. Fair enough, they're getting the worst of it. But it's affecting senior ones too, just more quietly. It's just as important to keep your tools sharp as it is to learn how to use them in the first place, and the way we used to sharpen them is dying.
The weekend project
Here's how it used to work. You'd pick something just outside your skill, something you didn't quite know how to build, and you'd noodle on it for a weekend or two. No deadline, no stakes, no budget. By the end you had some half-finished thing you'd probably never touch again, and a head full of stuff you didn't know two weeks ago.
It was a learning machine disguised as a hobby. And it was free.
That's the part of this vibe code epidemic that's been bothering me. It's killing that loop. Now you just pay to get a result, and the result doesn't teach you anything. The agent picked the framework, the agent fought the weird edge cases, the agent learned the lessons. You watched.
And notice what else happened: we swapped what was essentially a free hobby for one that's super expensive. The weekend project used to cost electricity and coffee. Now it costs real token money, and the thing you're buying is the removal of the only part that was valuable.
The rarest skill in software
Which leads to a weird conclusion. The person who gets the most out of these tools is whoever is best at resisting them. The ideal is to only reach for the agent when you wouldn't otherwise learn anything new, and in software that's an exceedingly rare situation. Almost everything teaches you something if you do it yourself.
So every time you hand a project to the agent, you're paying twice. Once in tokens, and once in the opportunity cost of the learning you skipped. We only ever talk about the first price. The second one is bigger.
Case in point
Say you spend 8 hours on a little web app for yourself. Nothing fancy. Maybe you pick a new framework, you noodle it out, and it's done. You may never open that app again. But six months later you might use that framework on a tool for your company that saves them a million dollars. That's the weekend project working as designed. The app was never the product. You were.
Now consider the pitch the other way. The agent builds it for you, and you got to spend your Saturday playing a video game instead. Sounds great until you do the math. All the tokens bought you was time, and you spent that time on the couch. In reality you just paid token prices to play a video game.
The new framework never made it into your head. The million dollar tool doesn't happen, or it happens slower, or somebody else builds it.
Keep some projects for yourself
I'm not saying never let the agent cook. There's plenty of work where I genuinely won't learn anything new, and that's exactly where these tools belong. But the weekend project was never about the output. It was the cheapest professional development program ever invented, and we're paying money to shut it down.
It's just not worth the opportunity cost to use this stuff all the time. Some projects need to stay yours, done the slow way, on purpose.
tl;dr: AI coding isn't just a junior dev problem. The weekend project used to be a free way for senior devs to keep their tools sharp, and vibe coding replaced it with an expensive way to learn nothing. Every delegated project costs you twice: the tokens, and the learning you skipped. The second price is the bigger one.