AI 正在推动适度成功的软件项目迎来爆发
原题AI is driving a boom in modestly successful software projects
读前速览
通过统计约 20 个软件市场的安装量，作者发现 AI 带来的是适度成功的业余项目（插件、模组、扩展）约 2 倍的增长，而非现象级软件的爆发。
AI 带来的是适度成功软件的可测量爆发，而非现象级产品。
在约二十个市场和注册平台中，作者发现跨过适度门槛的项目大约增长了 2 倍（例如下载量 1000+ 的 NVDA 插件、安装量 100+ 的 HACS 插件、评论数 100+ 的 Steam 游戏增长约 60%），而最顶端（下载量 10000+、安装量 1000+）的数量没有变化。
插件、模组和扩展比 AI 工具类仓库更能反映 AI 的真实影响。
它们公开安装量，因此使用情况可以被验证而非推测；而且它们大多解决的是普通用户的问题——在浏览器里玩《光环：战斗进化》、在《绝地潜兵》里开飞机、《任天堂明星大乱斗》的延迟回滚网络代码——而不是关于 AI 本身的软件，后者只是在「卖铲子」。
自愿披露 AI 使用的数据无法回答这个问题。
业余社区对承认使用 AI 抱有敌意，因此一个披露标签会抑制采用，许多开发者干脆不会承认；作者因此依赖总体趋势变化，而非自我报告的归因。
大型软件公司因结构性原因，AI 驱动的功能增长很少。
成功的产品已经拥有了他们想要的可见功能，被现有的功能面拖入技术瘫痪，而它们剩余的大量工作是微妙的（企业兼容性、内部可靠性系统），而非用户可见的。
没有出现极其成功的新业余项目，可能是时机问题或创意稀缺问题。
要么好的产品创意依然真的很难，AI 无法提供；要么成功的项目需要数年才能复利增长，这意味着今天拥有上千用户的工具可能需要耐心才能成为拥有十万用户的工具。
完整摘要
这篇文章回应了一个反复出现的问题：为什么 AI 编程智能体没有催生出大量优秀的新软件？作者从约二十个来源抓取了安装量和下载量数据——NVDA 插件、HACS 插件、Steam 游戏、Obsidian 插件、浏览器扩展、Nexus 模组、Blender 扩展——并对比了 2022-24 年与 2025-26 年之间成功项目的数量。他发现适度成功的项目大约增长了 2 倍（例如下载量 1000+ 的 NVDA 插件、安装量 100+ 的 HACS 插件、评论数 100+ 的 Steam 游戏增长约 60%），但在最顶端没有任何增长。他认为插件和模组是一个很好的观察窗口，因为它们本身不是 AI 工具，反映的是普通用户的需求；而自愿披露 AI 使用的数据不可用，因为业余社区会惩罚承认使用 AI 的人。他推测，大公司已经拥有了他们想要的功能，且在技术上陷入瘫痪；而极其成功的业余项目之所以罕见，是因为好的产品创意依然很难，且需要数年才能成熟。
目录 · 4 节
If AI is so great, why aren’t we seeing an explosion of new, amazing software products? If anyone can turn their idea into working software with coding agents, where is all that software? People have been asking this question for years. As far as I can tell, the answer is that it’s spread out over a lot of modestly successful projects:
- Somewhere between 2x and 20x new software is being created, depending on where you look
- There’s no real growth in massively successful software projects (not counting AI tools)
- There is around 2x growth in modestly successful software projects
Where all the new software is
I picked1 around twenty separate sources — software marketplaces, plugin registries, modding platforms, etc — which expose installation counts. I then compared the number2 of successful software projects between 2022-24 and 2025-26: i.e. before and after agentic coding hit the mainstream in mid-2025.
There’s a clear jump in modestly (not wildly) successful projects. For instance, the NVDA screen-reader addons directory had a bit over twice as many new addons with more than a thousand downloads (though no change in addons with more than ten thousand). The HACS open-source home automation community plugins store likewise saw twice as many new addons with over a hundred installs but no change in ones with over a thousand. New Steam games with over a hundred reviews are about 60% more common now than pre-2024. Obsidian plugins, Firefox and Chrome extensions, Nexus game mods, and Blender extensions had the same pattern. The Google Play store, WordPress plugins, and a few others didn’t show this pattern, but I think it’s still enough to conclude that AI is indeed driving a boom in hobbyist software.
Why addons and plugins?
I think examining addons, mods, and plugins is useful for a few reasons. First, it’s usually pretty easy to tell how many people are using them, or at least have installed them. Obviously AI is letting people produce a lot of software — I work at GitHub, so I see this first-hand, but the explosion in vibe-coded repos is clear to everyone — but without checking installation counts it’s unclear how much of this is actually being used.
Second, it’s a good way to minimize the amount of software that’s about AI. It doesn’t tell us much about AI’s effect on software to point out all the successful software projects that help people code with AI. That’s just selling shovels. The real question is whether AI is helping generate software that solves people’s ordinary problems.
AI-generated game mods and ports are a great example of this, because gamers aren’t trying to change the world with AI coding. They just want to play multiplayer Halo CE in the browser, or fly a plane in Helldivers, or play Super Smash online with rollback netcode. It’s looking like3 AI can supply this need effectively.
Were these projects actually AI generated?
Everyone is very coy about AI usage in hobbyist communities. In the rare cases where people admit it, they often get dogpiled. People really don’t like AI, even when it’s producing software they find useful. Given that, I don’t think it’s possible to use voluntary AI disclosure data to figure out whether AI is driving a software boom: the presence of the tag itself will discourage use, and many developers are just not going to admit it.
A 2x growth in successful projects between 2022-24 and 2025-26 is clearly something to do with AI, particularly in the background of a 10x growth in total projects. Of course not all of these projects will have AI involvement. But when people ask “why hasn’t AI improved software”, they’re talking about a missing trend4: wondering where all the new software is. As it turns out, it’s going towards a lot of small software projects that a few hundred or thousand people find useful.
Why small, moderately successful projects?
Why is AI growth visible in small projects but not big software companies? Why is it concentrated in moderately successful products and not wildly successful ones? I don’t have any data for this, but here’s my speculation.
Big, successful software companies often have all the visible features they want That’s why there’s not a wild explosion of new features in, say, macOS or Gmail: what would those features even be? As I’ve written about before, they’re also technically paralyzed by the features they have, so it’s not trivial for even very good coding agents to add new ones. Finally, many new features in successful products are subtle (e.g. enterprise compatibility things that almost all users will never encounter, or purely internal systems that slightly increase reliability).
It’s less clear to me why there aren’t many new wildly successful hobbyist projects. Maybe it’s just hard to have a really good idea for a product, and AI can’t help with that. Or maybe it takes years for a really successful project to develop, and we need to be patient for the current crop of thousand-user tools to grow into hundred-thousand-user tools.
-
As you might expect, this research was all done with the help of AI. It wasn’t purely vibed — I gave a lot of feedback about where and how to look — but I would definitely not have scraped or analyzed this much data on my own. If you want to see the actual research data (or more likely point an agent at it), I’ve pushed it here.
↩ -
Sometimes scaling to take into account platform growth. For instance, between the two windows I measured, Obsidian relaxed their requirement that all new plugins be reviewed by an admin, which predictably caused an explosion of plugins on the marketplace.
↩ -
It certainly seems like it can so far, but who knows. Maybe once the hype has died down people will return to less buggy artisanal human-made mods and ports.
↩ -
They might also be asking “why hasn’t the software I use gotten better?” For that, see my thoughts in the final section of this post.
↩
If you liked this post, consider subscribing to email updates about my new posts, or sharing it on Hacker News.
Here's a preview of a related post that shares tags with this one.
Software engineers don't get to be Luddites about AI
Many software engineers are righteously angry about AI. It’s understandable to be upset that the skills you built your professional identity around are getting automated away. But software engineers are the last people in the world who should be complaining about this. If you live by the sword, you will perish by the sword.
From the very beginning, software engineering has been about automating other people’s work. Every single commercial software program is valuable because it takes over some task you would have to pay a human to do. The engineer who built VisiCalc, one of the first pieces of personal commercial software, agonized over the potential of the spreadsheet to put bookkeepers out of work. Here is a brief list of industries software engineers have automated away:
Continue reading...