* Desktop first, no electron crap
* Open source and free
* Linux first
* Subjective, but to me it looks clean
If getting all that means using some AI vibe code, that's fine by me. Who isn't these days anyway? (Be honest!)
Anyway I hope the project is successful, more choice and competition in Matrix clients is a good thing.
Now if only they can fix video calls...
1. Forked an existing app that has all the features you listed.
2. Admit they don't really know the language and tech it uses.
3. Said upstream was too slow, but I don't see any(?) PRs from them on upstream besides the 2 that they list in the post. Their fork appears to have an extra 3,000 commits.
I'm not super against them doing this, but it's pretty easy to see why people don't like it. Hell, this is the same group that upgraded one of my side-projects from a few years ago and improved it into their 'baibot' matrix bot, so I wish them all the best. I like people making money from OSS, more power to them.
It made me think: we don't have to suffer the brokenness of the old bot (https://github.com/matrixgpt/matrix-chatgpt-bot) anymore. Still, I wanted something more thread-based and more powerful than what you had built.. and I wanted a playground to learn some Rust.
To clarify for anyone that might get confused: baibot (https://github.com/etkecc/baibot) is not based on any of the chaz code, nor on the matrix-chatgpt-bot code. It's completely manually-built / independently-built (in Rust), by me, over multiple months of unpaid FOSS work.
I can't say how frustrating it is to be midway through reading something and realise there's no human author.
Especially when it comes to the blog post, it's been human-reviewed and tweaked many times.
Still, if you don't like the content, it probably doesn't matter that there had been a human dedicating too much of his free time on this side project.
Source: the human author is me.
I also bounced off this piece around halfway through when I realized it was mixed AI/human content. I can read AI output anytime I want. Show me your true self! :)
Thanks for your work. Appreciate you!
Maybe it's because I scrolled down before reading, but I could instantly tell this entire text was AI slop from the massive overuse of emojis and bullet points.
There's no designer around to do a professional SVG logo, so we make use of what's there - AI to generate the (raster image) icon and then some tools to turn that into an SVG.
It turned out a little rough around some edges, but still good enough, I think. If anyone would like to polish it up, PRs are most welcome!
Icons may be a matter of taste. This one may be too detailed, but still scales well to small sizes. For now, I don't see why we should follow everyone else and go with an overly-simplistic icon.
As for the README emoji vomit: this may be a matter of taste as well. I find it makes it easier to scan through things, and I can assure you that each point on the README has been given much though and review by a human. Reducing it to "emoji vomit" is going too far.
Disclosure: I've been working on this Matrix chat app in my spare time over the past few months.
Icon can really be a generic SVG taken from some CC stock website, if there is no time for anything else, would look much better. Getting rid of emoji vomit is one prompt away. I might do a PR if I find free five minutes.
I do draw the line when they start putting them in console outputs though.
I really hope Komai start getting built for macOS.
When Show-ing your app on HN:
- Don't mention "vibe coding."
- Stay away from emojis.
Personally, I believe that a programmer's true skill today lies in how effectively they can leverage AI, and I’m all-in on vibe coding myself. But seeing the reaction here, I’m starting to think I should have thought twice before sharing my app on HN.
Even this post was translated by AI... I guess that’s probably a strike three, isn't it?
You still need skilled engineers to "operate" the AIs and to verify the results, but why on earth would you spend 3 weeks coding something by hand that Codex or Claude can spit out in a couple of hours.
For reference, at my employer we have so far in 2026 (1500+ developers) created 75% more code by AI this year compared to all of last year. Features are being delivered faster than ever, in a quality as good or better than before. As another advantage, we can now use skills to guarantee that the code that is created follows best practices, architecture patters, local governance and compliance. The "big thing" right now for us is providing guardrails and governance for agentic AI.
It is terrifying that any working software developer believes this is true.
LLMs are useful tools. But regardless of your inputs they guarantee nothing, ever.
It’s here and we need to write software. Let’s celebrate that it lets us do it better!
To my eye, the app is clean and minimal and shows me everything I need easily.
- Microsoft Teams icons. Really??? - Padding everywhere - Feels like a machine control panel, not a chat app - Double sidebars? Why?
This feels like a uTox clone. Not good in 2026. Even Teams has a significantly better interface.
Why double sidebars? Because Nheko established this pattern and I still think it's a good idea. On the desktop there's usually plenty of horizontal space to comfortably show everything without having to hide things away and require you to click multiple times. Still, both sidebars are collapsible to "just icons" for people who need the extra space.
In fact, most Matrix apps (Element, Discord, ..) have the same sort of "double sidebar" layout, but usually collapse the left-most sidebar to "icons only" by default. I think this makes it hard for new users to learn (icons are cryptic without a label) and there's no good reason to do it when you have enough horizontal space.
As for the padding, whether the defaults look good probably depends on screen size and personal preference. Komai is very configurable. There's a "Density" setting which offers 3 different options ("Spacious", "Compact" and "Dense"). "Spacious" is the default one, because it provides better ergonomics due to larger hit targets. Accessibility is important, but users who find the default padding too much can adjust it.
Disclosure: I've been working on Komai in my spare time for a few months now.
PRs for improvements are welcome, but:
- emojis are there intentionally. I asked for them and I personally think it's not too much. It's not something to clean up.
- I think the current AI-generated logo has character and is better than a stock SVG one can find online. The logo aims to represent a creature which is a mix between cat (nheko), lion & dog. See https://github.com/etkecc/komai/blob/main/docs/user-guide/id...
------
If you're choosing your software based on "has emojis in the README or not", then you'd be happy with many of the other Matrix clients. Most don't use emojis in their documentation (if they have much documentation at all).
In a software house with 1000+ developer, we're using AI everywhere. We have plenty of skilled engineers to validate the output of LLMs, but instead of spending 3 weeks developing a new feature, they can now spend a couple of hours or days.
We absolutely still need skill engineers for keeping track of what goes on, but their productivity has skyrocketed when they no longer have to obsess over details for a long time.
It's not like they woke up and went "I'm selfish, let's make everything a little bit worse for everyone" - but evil wins when good people vibe slop and we just let it go.
What a crazy era we live in where people can finally take full credit for someone else's work, because that someone else is an algorithm that even encourages you to do so.
Komai is explicitly presented as a fork of Nheko. We credit the upstream project, major libraries (matrix-sdk) and assets (iconsets, etc.), and are open about using AI during development.
I’m not claiming sole authorship or pretending Komai was created from scratch. It's built on existing work (done by various humans over many years) and we're being transparent and thankful about it.
Disclosure: I've spent multiple months working on Komai (with AI and other humans).