为什么我们还没看到大量新的好软件？AI 编程时代的软件改进为何不易察觉

Nate Meyvis

12 Oct, 2026

Here's a question we see, especially from AI skeptics: if AI coding is so great, shouldn't we be seeing more new, good software by now?

Here are some answers. As usual in this style of post, these are conjectural, candidate answers, and I would give some of them little weight.

We are in fact seeing software improvement, in the form of fewer bugs and better performance.
We are in fact seeing software improvement, in the form of more and better features in existing products.
We are in fact seeing software improvement, but the "we" is very limited. Most people don't adopt new software very often, so it's not surprising that they aren't finding new great software now.
Many businesses' software-development processes have subsystems, especially political subsystems, that have not been accelerated with AI. (Imagine the process of getting senior engineers and product managers to sign off on a development plan. The engineers and PMs could use AI to do this better and faster, but perhaps they aren't yet doing that.)
AI is replacing more durable, longer-lasting software with more ephemeral forms of software or software use, and one might not easily perceive this as "better software via AI." (If you do your taxes with Claude and don't bother building or shopping for better tax-preparation software, you don't quite get the "this product is like TurboTax but wonderful" experience, however much you might benefit.¹)
AI is replacing broadly available software with private software. Domain experts are building their own tools; executives are making their own dashboards; companies have bespoke internal projects. So the software is out there, just not visible to the rest of us, and especially not visible to skeptics.
We're making more software instead of making software better. We've always had the choice of (roughly) improving what exists or pushing development forward, and we've often chosen the latter. It's not surprising that we're still doing a lot of the latter.
Software teams are capturing a lot of the speedup in the form of more leisure or quasi-leisure time. (To greatly simplify: software teams are offering 2x speedups to their companies. They're then using tools that allow them to work 10x as efficiently, and taking nice long coffee breaks.)
AI improvements in software sometimes push us to use and make software less. Sometimes AI helps me figure out that a software project is a bad idea. Sometimes it nudges me to spend more analog time with humans. Sometimes it helps me find, and encourages me to read, more and better books. The net effect of this for me still involves my spending way more time making and using software, but it might be different for others. Most of the time, we're improving not just software but some system with both software and non-software components. If someone's Letterboxd (or similar) is working perfectly for them, they're likely to simply be a bit happier with their video-consumption life, potentially in ways that will crowd out their using other software.

Again, some of these are speculative and some should be given little weight.

This is just presented as an example, however realistic. It's certainly not tax advice and I am emphatically not a lawyer or tax professional or anything similar.↩

#generative AI #sociology of software #software