I killed tokenmaxxing
This blog post was released on the twenty-sixth of June 2026.
Big important news
I took my time to make a post, as I had some extremely immediate matters like applying for permanent residence. It was occupying my mind heavily and I couldn't really do anything else. Now I am way more relieved.
Anyway, I did it, so dear immigration officer, if you read this, everything on my application is true. I hope I made it as easy as possible for you to just see my application and say, "oh god it's so good we are giving you a pr right away." That's my plan.
More important news
Well, important for me, at least. I released an alpha of agentLIBRE. I think I will put more effort into docs, wiki, and making it more user-friendly later. Now, I make sure I have stuff to make user-friendly.
I am back to Zed; it has ACP Codex integration, which is interesting. No more pure vibes, I created a workflow where an LLM creates HTML docs with code review. Not perfect, but good enough for now.
I review 100% of my code with my eyes. If there is some slop/garbage/bug, it means that I missed it. I think it's better to miss something and take responsibility than for people to know that a product was vibed and developers have never seen that code with their eyes. That would be detrimental, wouldn't it?
Features, features and features?
Well, it's an alpha, so a lot of stuff will be broken on a regular basis. I think I will go full alpha-beta-release cycle. A lot of questions must be figured out first, and the best way is to try to implement and use them.
agentLIBRE-legacy is a horrifying monstrosity of tokenmaxxing. It became a dump of ideas, and now I am reasonably transferring ideas/features from this old repo. Still, there are many features to implement.
Stuff to do
I will create a separate web page for agentLIBRE at some point, as it not only deserves it, but it will also help with being user-friendly.
Maybe at some point I will create some adapter libraries to make it work on Windows, but for now I want to make a Linux/NixOS environment as seductive to the general population as possible.
I like shit talking Windows
I don't believe in Windows becoming better, and there is still no point to use it unless you are held hostage by the necessary software. I was there; I understand how difficult it is.
Sometimes I feel like I was extremely lucky to daily drive a MacBook, as macOS was enough for many developers to keep their distance from Windows' bad engineering, as well as to have a pretty UI with software support of Big Tech companies.
Story about scientific interest
I also remember having a conversation with one particular (today, it's called neurodivergent) person about the Windows networking stack.
That person said, "How cool it would be to figure out networking in Windows." I said, "Why would you bother even? You can examine networking in Linux freely." She answered, "But I want Windows, not Linux."
The point of this story was that my interlocutor wanted to examine something for the sake of examining. I had experience with WinAPI and Microsoft docs in general, and knew that even if you wanted, you would not succeed.
It stays true today. I highly doubt it, but maybe there is some magnificent software inside Windows; we will never know.
Pretty words to finish the post
GNU/Linux components might be ugly from the outside, but beautiful in their openness. At the same time, FOSS starts being pretty from the outside, too.