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.
This applies to all software. You didn't make the monitor your app is being displayed on. Stop taking credit for the compiler that made your code runnable, etc.
There's a great difference between making a table with tools other people made, and ordering a custom table on Amazon, waxing it, and claiming you made it. This shouldn't even be an argument.
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.
If you mean Komai specifically, I think that this characterization is unfair.
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).
I was specifically addressing my parent comment, which was equating taking credit for LLM output to taking credit for making something using tooling or hardware.
Who cares if they used AI assistance? Vision is theirs, prompting is theirs, guidance is theirs, verification and iteration and feedback and so on is theirs. It's not like they zero-shot "make a Matrix client for Linux" and then just posted that with zero processing, review, testing, or anything. Sheesh.
Yeah! The burning of litres of waters is theirs, the degredation of their intellect and ability is theirs, the financial support of the companies ruining the internet and the world is theirs, the support of the hype machine ruining the economy is theirs!
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.
Let's not blame the victim for this. I know it's unorthodox to think that evil corporations are exploiting their users in order to get away with evil things.
I really like the Matrix ecosystem, and this client seems like a cool addition, but how are posts like this with a load of AI generated prose getting on the front page?
I can't say how frustrating it is to be midway through reading something and realise there's no human author.
While AI was used for writing the code and large parts of the text (including in the blog post), I can assure you there's been a human author dedicating many days and sleepless nights on this alongside the AI.
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.
I love matrix-docker-ansible-deploy and use it for my homeservers.
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! :)
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.
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?
Everybody is AI coding these days, and if you're not, you may find yourself out of a job soon.
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.
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.
Having the core of your app be written in languages you self-admittedly don't understand is a bold move. I've been a big fan of the ansible-matrix playbooks for a while now so I'm willing to see this play out, but it doesn't fill me with confidence.
Personally, I think the client looks cool, but the problem with Matrix has never been client-side, it's more like the entire protocol is bad at doing what it aims to do.
While I could understand some AI assistance, I just cannot look at such README with clearly sloppy too detailed nonsense app icon and eyesore emoji vomit:
This is a FOSS project with absolutely no funding.
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.
I do appreciate doing unpaid volunteer work. My comment sounds clingy, not is not a demand more of a pointing out.
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.
One person's "emoji vomit" classification is another person's "this looks good to me".
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).
It is utterly annoying to see emojis pollute information presented in a technical manner. It’s the hallmark of generated technical slop and seems to appeal to the perpetual resume polishers found on LinkedIn.
The icons are not from Microsoft Teams. It's the Fluent System Icons (https://github.com/microsoft/fluentui-system-icons) free iconset by Microsoft, which Nheko (Komai's predecessor) also used and we continue to use. Because we found this iconset lacking in certain aspects, in Komai we also supplement it with Font Awesome.
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.
Wow, so much negativity, when the app scores really high on things that are supposed to be important here:
* 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...
It's because they:
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.
Your Rust-based AI chatbot (https://github.com/arcuru/chaz) was an inspiration!
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.
People are afraid and taking it out on others
Vibe coding isn't bad. It's those people who do all there coding(?) with it and post it like it's there creation is the problem.
This applies to all software. You didn't make the monitor your app is being displayed on. Stop taking credit for the compiler that made your code runnable, etc.
There's a great difference between making a table with tools other people made, and ordering a custom table on Amazon, waxing it, and claiming you made it. This shouldn't even be an argument.
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.
If you mean Komai specifically, I think that this characterization is unfair.
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).
I was specifically addressing my parent comment, which was equating taking credit for LLM output to taking credit for making something using tooling or hardware.
Who cares if they used AI assistance? Vision is theirs, prompting is theirs, guidance is theirs, verification and iteration and feedback and so on is theirs. It's not like they zero-shot "make a Matrix client for Linux" and then just posted that with zero processing, review, testing, or anything. Sheesh.
Yeah! The burning of litres of waters is theirs, the degredation of their intellect and ability is theirs, the financial support of the companies ruining the internet and the world is theirs, the support of the hype machine ruining the economy is theirs!
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.
Let's not blame the victim for this. I know it's unorthodox to think that evil corporations are exploiting their users in order to get away with evil things.
I really like the Matrix ecosystem, and this client seems like a cool addition, but how are posts like this with a load of AI generated prose getting on the front page?
I can't say how frustrating it is to be midway through reading something and realise there's no human author.
While AI was used for writing the code and large parts of the text (including in the blog post), I can assure you there's been a human author dedicating many days and sleepless nights on this alongside the AI.
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 love matrix-docker-ansible-deploy and use it for my homeservers.
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!
> midway through reading something
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.
You let yourself be blinded. There’s nothing slop about this. Every sentence is crisp and meaningful.
It's great that etke.cc chose to build on nheko (a very fast Matrix client) and put in many thoughtful upgrades.
I really hope Komai start getting built for macOS.
Today’s lesson learned:
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?
Everybody is AI coding these days, and if you're not, you may find yourself out of a job soon.
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.
> we can now use skills to guarantee that the code that is created follows best practices, architecture patters, local governance and compliance.
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.
I'm merely reflecting on what reality is showing.
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.
Having the core of your app be written in languages you self-admittedly don't understand is a bold move. I've been a big fan of the ansible-matrix playbooks for a while now so I'm willing to see this play out, but it doesn't fill me with confidence.
Personally, I think the client looks cool, but the problem with Matrix has never been client-side, it's more like the entire protocol is bad at doing what it aims to do.
I’m glad that some people in the FOSS community are embracing AI assistance.
It’s here and we need to write software. Let’s celebrate that it lets us do it better!
I have been using it for the last 2 days, it feels blazing fast and I like all the appearance customization options
This looks very ugly compared to other matrix clients
This may be a matter of taste, because I quite like the approach the ekte team have taken.
To my eye, the app is clean and minimal and shows me everything I need easily.
Fair enough!
While I could understand some AI assistance, I just cannot look at such README with clearly sloppy too detailed nonsense app icon and eyesore emoji vomit:
https://github.com/etkecc/komai
This is a FOSS project with absolutely no funding.
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.
I do appreciate doing unpaid volunteer work. My comment sounds clingy, not is not a demand more of a pointing out.
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.
One person's "emoji vomit" classification is another person's "this looks good to me".
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).
It is utterly annoying to see emojis pollute information presented in a technical manner. It’s the hallmark of generated technical slop and seems to appeal to the perpetual resume polishers found on LinkedIn.
A lot of younger developers seem to really like this. Not a big fan personally, but you gotta live with the times I suppose.
I do draw the line when they start putting them in console outputs though.
Emojis can be nice, their color alone makes looking up headings easier. But everything has a limit.
C++ seems like an odd choice for developing desktop apps in 2026, but I guess if it works, it works.
What an awful UI.
- 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.
The icons are not from Microsoft Teams. It's the Fluent System Icons (https://github.com/microsoft/fluentui-system-icons) free iconset by Microsoft, which Nheko (Komai's predecessor) also used and we continue to use. Because we found this iconset lacking in certain aspects, in Komai we also supplement it with Font Awesome.
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.
love that pricing page
AI slop app.
You can teach a person to write programs. You can't teach good taste.