I'm not an expert, but something rubs me the wrong way about slapping TM on kernel, a word that already has a common technical use.<p>It would be like General Mills advertising its new breakfast(TM) cereal.
Looks like there's a demo, but I don't know what it's supposed to do: <a href="https://demo.rabbit.tech/webapp/en" rel="nofollow noreferrer">https://demo.rabbit.tech/webapp/en</a><p>Seems like I'm just talking to LLM and that's it?
Lots of void text, lots of trademarks (even for the word kernel™...), lots of smoke, and nothing to show without giving out your email address first.<p>I tried the demo. Most of the time the model understands the request but fails to deliver a valid response.
If you ask it to play a song, you'll get a Spotify login prompt running in a remote Chromium instance. The Spotify login is interesting... you can try to log in with google, then click the Privacy link at the bottom, scroll to the footer and click the Google link, then search for an online visual studio code and open up some files on the VM that they're running...
Personally, I feel like I can direct a computer to do what I want way more precisely using a terminal (or even classic GUI) right now than I ever could using the ambiguity of natural language.
I don't understand this at all but the demo was very disappointing. Basically just a Siri-style AI to talk to that can schedule tasks. Not very inspiring
lost me at "quantum engine".
it's a bunch of automations linked together by something like GPT with function calling.
and you'd have to give them access to your accounts to run things in "tHE cLOuD".
i asked it to play an album and it started streaming me the screen from a remote ubuntu instance asking for my spotify credentials -- press command+shift+i to bring up the window menu
I find the idea of putting a natural language model in front of my OS silly and misguided, because I naturally know how to use my computer and translate my thoughts into actions on the mouse and keyboard swiftly. Putting a (probably poor quality) text input as a barrier feels like it would degrade my experience
Bullshit artists trying to build vague techbro hype with the right buzzwords to hopefully attract enough Silicon Valley VC funding so they can turn up no product or a really hilariously bad product and walk away with all the money.
They don't lay it out clearly but it seems that they're aiming for a mobile device that allows you to dispatch tasks that are done in the cloud using natural language.[1]<p>I think this is exciting, it gives me vibes of LCARS and voice commands from Star Trek. Whether it turns into something more than hype we'll have to see.<p>[1] <a href="https://www.rabbit.tech/rabbit-os" rel="nofollow noreferrer">https://www.rabbit.tech/rabbit-os</a>
I love how asked about weather it told me it doesn't have access to real-time weather, so I then asked what it can do, and it replied with weather forecast I asked for but couldn't read because immediately it replaced it with "I can do whatever", i.e. the reply to the second question. :D
It seems like a fun idea, though I can't see any situation where it'd be useful to me, whatever it is. But then you add on the marketing, and it just seems like a sketchy startup. The TM mark on kernel (and reusing the term kernel in the first place) immediately raises red flags.
I want to balance the overall negative response here. I think people's criticism falls into one of the following categories:<p>- "This is AI bullshit.":
I haven't tested it extensively, but I think giving LLM's or similar constructs the ability to execute tasks and perform actions is the next step in AI capabilities. So this makes sense to me.<p>- "This is not an Operating System."
Maybe we need to rethink what an Operating System is. As long as computers were boxes under our desk, an operating system was the thing that operated these boxes. Now computers are everywhere and - when connected to each other in the cloud - form sort of higher level computers. Calling a system that operates the higher level computer an OS is legitimate in my eyes.<p>- "They put TM behind Kernel"
Ok, that one is weird.
Service as a software substitute. Weak. Imagine the world where this nonsense is taken to its logical conclusion. I am really interested in trying to use fifos / named pipes for interprocess communication with a locally running llm piping into a terminal. The llm would be trained to expect a bash terminal's output while attempting to accomplish some goal. Its outputs would be bash commands. The bash terminal would be piped to be the input of the llm. Circular loop. Other than security, I'm curious why I haven't seen any kind of project like this for automating autonomous agents. You could run it in a VM on the first computer. Is anyone aware of something like that, that isn't vaporware? I don't want a black box OS. I want an OS with a highly skilled pair programmer in-line with me at a terminal that I can call up when I get curious or want to know the smart person way to fix something. As a self taught computer person who does software as a hobby, something like that would be really attractive.