One of the weirdest things of the last six months was that software, which used to be extremely valuable, has started to feel…disposable. Accepting that idea is very, very hard. It goes against every instinct I’ve built over my career. My instincts tell me that software is difficult to make and valuable. What if my instincts are wrong?

I’ve got 20 or 30 vibe-coded GitHub projects that I created with great intentions, and they actually work—but they didn’t go anywhere. Now what? It’s weird to erase things in 2026; we tend to let everything sit there on the hard drive or in the cloud, gathering dust. I could open-source them, but that’s fake virtue; I have no intention of supporting them or fixing bugs. They’re slop products. Appslop.

The sudden cheapness of code has other effects. You talk with a team about what to build, they go away, and everyone comes back the next day with their own version of the product. They have that gleam in their eye: They solved it! I’ve felt it, too.

It used to be that everyone would talk for a week: Design would sketch, engineers would write on graph paper, product managers would draw on the whiteboard. People would go make various things, and a rough alpha version would emerge by bringing them together. But now, they can all go back on their own and make a version of everything, right away.

It’s fun to have options, of course. But this is one of those places where I can bring some knowledge from my career as a writer/editor: First drafts mean nothing. They’re hard to create, they feel incredibly valuable, and you may be very proud of yourself for getting it done. It’s a glorious feeling—but it is almost always totally wrong.

The whole culture of writing is filled with violent metaphors about this: “Kill your darlings,” for example, and lots of stuff about slashing and cutting and murdering. Editors work with red ink, to drive it home; they are often described as “wielding” a red pen. The way a professional editor treats a first draft is roughly how the Roman Senators treated Caesar. 

There’s a really great book by the truly excellent journalist Tracy Kidder, who just passed away in March. He wrote a number of amazing, carefully-reported books, but there’s one that’s more of a writing manual called Good Prose. Not many people could get away with that title, but he absolutely could. 

Critically, he didn’t write the book alone: It was a collaboration with Richard Todd, the executive editor at The Atlantic. The book isn’t about how to be a great writer; it’s about the process of writing and iterating together until prose gets very, very good.

Todd died in 2019. Kidder wrote then in the New York Times:

I can’t imagine that he ever lied to a writer about a work in progress, and in my experience, he was never less than astute about the writing’s flaws. And yet his criticism was never harsh, never mean—that is, never harmful. He was, in a word, altruistic. He worked hard and anonymously in the best interests of other people’s creations. It reminds me of something Paul Farmer, the subject of my book “Mountains Beyond Mountains,” said. Farmer used to tell his medical students that to be a good doctor, you must never let the patient know that you’re in a hurry or that you have problems, too. Todd lived that way for us.

The humanist-tech types among us have been saying that software is a medium for decades now. Nick Montfort described it thus in 1995; Alan Kay in 1984 said it was “the first metamedium, and as such it has degrees of freedom for representation and expression never before encountered and as yet barely investigated.” 

That’s all good stuff—but if you work in software, it’s often more like an industrial process than a humanist endeavor. You’re managing using Gantt charts or Kanban boards. Work is divided, assembled, tested, and there are lots of “assembly line” metaphors. In the world of vibe coding, people use terms like “dark factory” to describe prompt-driven software creation where it’s all bots, no craft. Dark factories sound gloomy, but on the plus side they’re less violent.

But wait a minute: What if 1995 was right? What if software really can be a medium now? Or rather, in order to vibe code, I literally just…write things. I have to know what I’m writing, but I tell the robot what to do in sentences. To fix code, I…write more things. Right now, way, way more code is being created this way than the old way, which involved, you know, actually writing code. 

When software people talk about this, they often say: “This is a long-term maintenance problem.” My guess is most of the stuff being created will simply be forgotten. We probably just need to get better at archiving things and throwing them away. I know it seems that vibe coding can “one shot” whole apps, but trust me, you will find over time that they are not good. They’re only valuable in that they can show you what to improve, like any first draft.

I wonder if the process of shipping software—especially product management—will, in the future, look more like the editorial process. Less about marshaling resources for the long haul, and less about coordination. More about getting a bright-enough person to describe a problem and turn it into software, and then getting them to turn it in—in its ugliest form—for discussion and feedback. I like thinking about the Richard Todd school of product management, working “hard and anonymously in the best interests of other people’s creations.” 

If this happens, it will be a very confusing change. The tech industry’s worst nightmare is to find that its huge industrial systems, which cost billions to build, are now more like little magazines. Imagine if you went to Apple’s giant spaceship campus and you walked through the front door of the main reception area and it turned out to be the offices of n+1? Nonetheless, my long-term bet is that software development will look somewhat less like a factory and more like making a magazine, or a film, or producing an album. Talk is cheap, prose is hardly more expensive, and now code is cheap, too.

Nothing will change right away—we’re still sifting through the wreckage of the old way of doing things. That said, when I think about the work I’d like to do over the next ten years in this industry, “Software Editor” feels like a pretty good gig. Maybe it’s something to look forward to.