AI and Projects That Deserve an Update

A few years ago I wrote about rebuilding this website. Since then, pretty much everything has gone stale. Some of the dependencies were practically dead, the todo list was gathering dust, and the whole project was slowly turning into exactly the kind of project you'd rather not touch, because you know exactly how painful updating it is going to be.

What used to take weeks

Refactors like this always carry a certain mental weight. You push them back for months because you know that eventually you'll have to spend a couple of weekends carefully making changes, dealing with broken builds, and doing the same boring cleanup over and over.

But all of that was before the world got blown up by AI. Things are different now! Not the "press a button, get a startup" kind of different that the internet loves talking about, it was more like sitting next to a very fast junior developer who never gets tired. I'd explain what I wanted, let it handle the repetitive parts, review the output, fix it where needed, and move on.

What used to take weeks got done in a single day!

Finally adding a projects page

This was one of those things that had been on my list since the very first version.

For a long time now, I've been writing way more code than I've been writing words. None of that work was visible on the site. The obvious fix was to add a projects page. The less obvious problem was maintaining it.

A projects page goes stale fast.

You build it carefully, get everything just right, and six months later half of it is stuff you don't actually care about anymore. Updating it becomes a chore, and slowly you stop bothering.

So I decided not to maintain it at all. Just pull the list of public projects straight from my GitHub account, that way, when I build something new or update an existing one, it shows up on the site on its own.

No forgotten updates, no "I'll clean it up later", no hassle.

The bilingual problem

Well, not entirely no hassle. The site has been bilingual for a long time now, English and Farsi. That's not a problem for articles, because I write them in both languages from the start. That's the whole idea. But projects are different!

Everything on GitHub is in English: the repo names, descriptions, all of it. The result was that the site would switch cleanly between the two languages, until you hit the projects section, and suddenly everything was in English again.

Enter AI

This is exactly the kind of thing an LLM should be able to handle easily, right?

Now the Farsi version of the site translates the project descriptions on the fly. The descriptions on GitHub stay in English, but when someone opens the Farsi version, they see clean, natural Farsi.

The translations aren't just word-for-word. Technical terms still read the way they should. The phrasing feels natural. The model knows where it shouldn't translate something directly. The final output is closer to what an actual Farsi-speaking developer would write about their own project than what you'd usually get out of machine translation.

And more importantly, I don't have to think about it anymore.

The bigger shift

None of these changes are huge on their own. At the end of the day this is still just a personal website and a simple open source project.

But rebuilding it made one thing clear to me, AI can rescue forgotten projects.

Slowly, the real bottleneck isn't the effort of building things anymore. It's figuring out what's actually worth building in the first place.

PS:

As always, the code is up on GitHub under the MIT license, go nuts with it!