One Year Of Gitus; Random Thoughts


If you don't know what Gitus is, it's basically like Gitea/Forgejo but with much less stuff - you host and manage Git repositories on it. I started working on it since late March 2025 for two reasons: (1) I want to learn how Git works internally, and (2) I hate Drew DeVault (mostly for his smear campaign against RMS, but he's also an asshole in general) so I've decided that SourceHut is not to be trusted. About three and a half months later I felt it's good enough to be tagged with version 0.0.1, and about two months later I felt confident enough to post this on Hacker News - twice. Turns out no one wanted a new Git repository hosting software in 2025. I understand it's very immature even now (I just fixed like two dozens of security issues a few days ago), but it do feel pretty unfair when someone did the same thing literally gets 100x the attention than I do - they didn't even have proper SSH support when Gitus supported it since the first release! Somehow my Mercury tutorial gathered more interests from people than this. Ridiculous.

Now I suppose if most single devs were in my position the most likely sentiment would probably be "I think it's done; I've did a good job here and it's time for me to move on" and they probably would move on, but I've decided not to do that. In the beginning of developing Gitus I've made a decision to only work on one side project for every single year instead of working on one for one or two months and then abandoning it like what I've did with all the programming languages I've designed and implemented before this. Naturally even I don't strictly follow this rule myself, but I am confident enough to say, that within the last however-many-months since March 2025, this is one of the biggest things in my life.

I still want to believe that I've made the right choice with that, that sticking to one thing and perfecting it is actually worth something. After all, I am easily distracted, Gitus do need long-time maintaining regardless of having actual users or not, and this would become some kind of training now that most of the "natural" hyperfocus I can conjure up in my younger days seems to have disappeared. When people started talking about how "in the age of AI" "people with a bit ADHD would come out ahead" I can't help but cringe; somehow they think having ADHD means your brain has extra capacity and can naturally manage multiple agents working on different projects with different context at the same time and that will get you "ahead", I had this strange irritating feeling like how your compass has came out of its box with its needle poking through your backpack so on your entire trip from school to home you got randomly pricked by the compass on the back. I can't say for other people, but to me, having to force yourself to switch between many different contexts is just not a good state to be in; it's either you fry your brain eventually or you have to give up understanding what's going on, and I don't want either of them.

I still write the code for Gitus myself instead of relying on a coding agent because now I also treat Gitus as a training gym so that I won't forget how to actually write programs. To be honest I'm not completely against LLMs and coding agents - they've proven to be a good learning tool, I don't make a it a point of how immoral it is and how it's paining me when I prompt in my day job, and it did give me very useful info when I use it for security audits, but I did feel the joy (and more importantly, the meaning) draining away when I was fixing the bugs a few days earlier when your brain can't help but keep reminding yourself that whatever you're doing right now is just one prompt and several minutes away.

I have what it takes, but the world seems to not want them anymore. Don't glorify this; Lord knows there's no heroism to this.


Back

Last update: 2026.7.14