Why I’ve Never Written a Single Piece of “AI Software” — and My View on It

Welcome to the post-ChatGPT era, where nobody cares about solving real problems anymore. Everyone is just trying to add AI to everything.

But I, an AI enthusiast, have never written a single piece of AI software, despite having done a lot of research on AI.

Let’s begin with something fun by looking at what Microsoft has been doing.

A random guy named Tey Bannerman figured out that Microsoft (a.k.a Macrohard) has named more than 80 distinct products “Copilot.” Yet we don’t understand why, since thanks to Macrohard’s effort, nobody’s using their AI product anymore — including the once-famous GitHub Copilot. Cursor and Claude Code have eaten much of its mindshare.

(Go to https://teybannerman.com/strategy/2026/03/31/how-many-microsoft-copilot-are-there.html if you’re interested. Tey even drew an interactive map for this — how ironic!)

Yes, I apologize that I have just wasted a whole minute of your precious life, and even more if you were trapped by the rabbithole.

Now back to our old friend AI.

We have been caught between the excitement that AI will become as powerful as human beings and the fear that anyone not “into AI” will lose the race.

And this is completely wrong.

First, scaling laws are running into bottlenecks, and thanks to the AI slop generated by all kinds of people, we don’t have any additional high-quality training data anymore, which makes the task even harder. As former Meta Chief AI Scientist Yann LeCun says, “We’re actually very far from [AGI].”

Second, not everything needs AI. If we need AI in everything, then we should pray for the rapture to happen immediately, and the world will still run normally since the AIs will do all the work. AIs are here to serve and facilitate us, not to replace us.

For example: Folo is an RSS app that I’ve long favored. It is created by the god of RSS, DIYgod, the creator of RSSHub. I enjoyed it when it served simply as the frontend of RSSHub — it felt like Inoreader combined with Reeder, with cross-platform support added.

But things changed. It started to introduce a paid subscription, and basically an AI plan. At $49.99 per year, you receive 30 AI summaries and 300 AI translations every day.

This is the exact opposite of why people still use RSS, and the exact opposite of how to build software. We escaped from Twitter and other social media platforms because we had enough of being flooded with information every day. We had had enough of algorithms that knew us better than we knew ourselves, controlled our emotions, shaped our political views, and did all of it for profit.

We subscribe to RSS not only to fulfill our own reading needs, but also to support creators who continue to produce high-quality original content — things actually worth reading. We choose exactly what we want to read, and we will probably read every new feed because we know exactly what we want.

So what is the point of letting AI summarize it?

Even if so, what’s the difference between using that and copying and pasting it into ChatGPT, while the latter costs nothing, and the former is fifty bucks?

I’m not against integrating AI into existing software. However, 99% of the “AI Software” (and startups, too) are still ChatGPT wrappers, and they’re CREATING a use case for the user that even the product managers who designed them won’t use.

The criteria are simple: if you ain’t gonna use this software, don’t build it.

Take Folo, for example. Introducing AI summaries was a terrible idea. But introducing an algorithm to browse the internet and find sites with fewer visitors and high-quality content suited to the user’s needs is exactly what users want. It solves a real problem for the user: the higher the quality, the harder it is to find.

Another example is my own video downloading software, LinkDown. Introducing AI to it is unnecessary — why would a user want a chatbox in a video downloader? But this was exactly what nearly everyone was doing in the past few years.

We’ve seen AI in to-do lists, bank apps, telecom operator apps, messaging apps, PDF apps, delivery apps, and much more. When will this come to an end? It’s time to stop this.