justin․searls․co

I've got your shovelware right here

Mike Judge thinks developers may be deluding themselves into thinking AI coding tools are making them more productive:

My argument: If so many developers are so extraordinarily productive using these tools, where is the flood of shovelware? We should be seeing apps of all shapes and sizes, video games, new websites, mobile apps, software-as-a-service apps — we should be drowning in choice. We should be in the middle of an indie software revolution. We should be seeing 10,000 Tetris clones on Steam.

If you read the post, he's got data to this effect.

I've got two problems with this:

  1. Much of his data basically stops around March–April 2025—the precise window when AI coding tools finally became worth a damn

  2. Whether or not we're seeing a shovelware boom is orthogonal to the question of whether canny developers' output is being supercharged by AI. All it takes is one developer to show one project that was delivered in less time than they'd have been able to do otherwise and bingo-bango: AI tools as they exist today—flawed as they might be—can do something a human couldn't have done otherwise. Sure, you could make an argument about the macro-level effects—like whether the vast majority of professional programmers are too stupid or resistant to change to leverage these tools and are even somehow being slowed down by their existence, but I'd probably agree with you

Seriously, suggesting that AI-generated code is a nothingburger because we haven't yet been drowned in shovelware just four months after coding agents became remotely useful? Get outta here. And right now, only the early adopters are even using them! I talked to a manager the other day whose team has been given carte blanche to burn through all the Anthropic tokens they want and for whom not a single developer touched the account in the month of August.

I think it's important context to know that a lot of the statistics Microsoft and Anthropic put out are inflated because they have an obvious conflict of interest to push the message that the thing they sell is useful. I'm sure because many developers have GitHub Copilot turned on, they tend to press Tab without reading, then waste time mostly deleting what it produces. But, because the user mindlessly pressed Tab, somewhere a cell in a spreadsheet labeled "acceptance rate" ticks up. In practice, I suspect far fewer people are using these tools in anger than the big tech companies are incentivized to portray.

Having a full-fledged agent that can build shit for you and verify things work is—when it comes to productivity—nowhere near the same league as tacking autocomplete and an LLM chat sidebar into an editor. And here's the promised counterexample: I built some shovelware of my own last week. Took about a day of wall time to build and another calendar day for me to tighten up with feedback. I spent probably a grand total of three hours staring at computers in the furtherance of the project. This blog post about it took me twice as much time. Would have taken me weeks to build by hand, and more importantly, it wouldn't have been built at all—I wouldn't have bothered.


Got a taste for hot, fresh takes?

Then you're in luck, because you'll pay $0 for my 2¢ when you subscribe to my work, whether via RSS or your favorite social network.

I also have a monthly newsletter where I write high-tempo, thought-provoking essays about life, in case that's more your speed:

And if you'd rather give your eyes a rest and your ears a workout, might I suggest my long-form solo podcast, Breaking Change? Odds are, you haven't heard anything quite like it.