Post

Once again, I'm about to talk a little bit about AI and how I've found it useful. I know some people think all AI is bad and there can be no good uses of it. Good for you if that's true. Please just ignore, mute, or block, rather than lecturing me. Now...

... for years I've had an idea for a kind of neat tool for which there's nothing really on the market (to be clear, it's a tool that many people would argue there are services that accomplish the same thing, but NOT THE WAY I THINK those tools could or should work for a specific use case).

I've been using a few services that get me part of the way there, it's always a kludge, and the one company that has the closest such tool is focused on a very different market such that I'm basically using their tool in an "off label" way to only kinda sorta get what I want out of it.

(I don't want to get into the details about what kind of tool, because then I know my mentions will fill up with people going "just use [blank] and shut up"). The tool I *want* felt like it would be somewhat complex to build. There's a lot to it. But last week I decided to see if Claude Code could.

I wrote a fairly long PRD explaining what it was and had Claude Code go into plan mode. We had to make a few decisions back and forth but it laid out a fairly light weight way to build it, showing how we could build it modularly to test things and experiment with a goal towards eventually using it

for real. I put the plan aside all week and then last night said "ok, let's start." And within 30 minutes it had all the basic feature working (which was insane to me). Then today, while I was outside doing yard work (but checking in via my phone) it basically built the rest.

It's very barebones and needs polish, but in a very short period of time I appear to have built a tool that I've wanted for years. Yes, there's still some work to make it truly production ready, but it's kind of crazy.

Again, I know that some people will insist that there are tools out there that accomplish what I want. And there are, but only at a superficial level. There are some specific unique features that make all the difference in the world to me, and no one else is likely to build them.

Yet, I've been able to build it for myself in almost no time at all. This ability to instantly customize things to *your* specific needs, rather than the needs of the company providing the service is such a big change. It's kind of crazy what possibilities it starts to open up.

Also, one other thing to be clear: this is a tool that is for (mostly) internal use (there may be *some* occasions where I will expose part of it for others to use), but is not something I'm building for others. If I were building it for others, I think there would be a lot more work. But still...

I still don't think people insisting this stuff is useless recognize how far the coding features have come and what it opens up in possibilities for those who want more control over their own tools and who don't want to wait for someone else to build what they want.

And, yes, I know that making software is one thing, maintaining it is another, and maybe I'm in for a huge surprise down the road. But again, I'm a bit less worried about that given that I'm not really supporting other users. Just me. And maybe a couple colleagues.

Anyway, just pointing out that the realm of what's possible keeps expanding for me, and I think that's important to recognize. For all the complaints about enshittification, the coding tools open up a really interesting alternative. /end

1 repost34 likes

It really is wild what someone capable of self-hosting an app can accomplish. I am very curious to see if there's enough demand for the models capable of what you've described to be run profitably so they remain available.

This is a good example why I think news of the demise of IT is premature. AI is going to make the number of specialized tools explode and a lot more companies, or even individuals, will be building exactly what they need in house.

Just the pro plan, not max. I actually thought that this project would definitely take a while because of bumping up against limits, but somehow... it hasn't. I'm kinda shocked. In the past I've bumped up against the limits and just had to wait. But this one (which I thought would be a lot) was easy