Inquiring Minds Want To Know...
On what hardware is this being tested? It IS being tested, right?
Torvalds challenged the haters to fork Linux. Someone said 'hold my beer'
Earlier this week, Linux project leader Linus Torvalds told AI haters to fork off, and invited anyone who didn't like his comments to fork the kernel. Well, here you go: linux-0.11-rs, a total reimplementation of the Linux kernel, done in langage de programmation du jour, Rust. No, this isn't really a response to the Emperor …
-
-
Friday 17th July 2026 15:21 GMT Liam Proven
Re: Inquiring Minds Want To Know...
> On what hardware is this being tested?
It is right there in the Github README. You did get that far, right?
The VERY FIRST LINE says:
"A modern Rust rewrite of the Linux 0.11 kernel — boots on i386 in QEMU, runs a self-hosted Unix-style userland."
> It IS being tested, right?
It's a toy. It does not matter if it works past booting in a VM.
-
-
Monday 20th July 2026 14:39 GMT dlc.usa
Re: Inquiring Minds Want To Know...
No, I did not get that far. I have relearned the need to follow every link before commenting on an article. My sincere apologies for assuming from the article proper no testing had been attempted. I am very very glad to hear it was and has been successful--that is huge!
-
-
-
-
Friday 17th July 2026 17:13 GMT Anonymous Coward
Err? Why do we need AI in a Kernel?
We have managed NOT to have it for decades. Around 60 years at least. If we must have it, then what does it bring to the party?
The kernel is running before the network stack is setup so remote LLM'S are out of the question. Can the AI enabled Kernel fit in the small partitions usually allocated for kernels? (/boot)
We need to know these things.
Until then, FSCK AI and all it stands for. There is nothing Artificial or Intelligent about current AI models.
-
Friday 17th July 2026 17:13 GMT Anonymous Coward
Please Be Careful With Those Comments.....
.....in 1996 the Chinese economy was absolutely nowhere.................
....but since then it is SEVENTEEN times the the size.............
Over the same period the UK economy is only (maybe) THREE times the size............
Yup......"growth" comes in two types......the Chinese type and the (smaller) UK type......
-
-
Sunday 19th July 2026 13:01 GMT Anonymous Coward
Re: Please Be Careful With Those Comments.....
"Meaningless"??
So....over some period your income increases SEVENTEEN times......while........
.....your close buddy only has their income increase THREE times.
Do you console you buddy with "No probs.....it's all meaningless!".......or what?
-
-
Friday 17th July 2026 17:50 GMT Rol
Re: Please Be Careful With Those Comments.....
Depends how you measure growth.
In the UK, growth has largely come from an increase in population, while keeping a foot on the neck of house building. Take away the meteoric rise in property values and eye watering rents, and you're probably left with something close to zero growth in the UK.
-
Saturday 18th July 2026 10:49 GMT Caspian Prince
Re: Please Be Careful With Those Comments.....
Also depends how you value growth, as in, if you value growth above all else, great but a bit of a shame it came at the total collapse of the environment that support all life on Earth currently. As for myself I believe that growth as an end in itself is an utterly disastrous metric for success by any measure.
-
Saturday 18th July 2026 16:22 GMT Rol
Re: Please Be Careful With Those Comments.....
Entirely agree with you. Growth, often as not feeds inflation, which slowly chips away any semblance of fairness in wages. If I could be on the same proportionate wage to my boss today as I was in the 80's, I'd be buying Ferrari's for fun, instead of facing homelessness in two weeks.
-
Sunday 19th July 2026 21:24 GMT Anonymous Coward
Re: Please Be Careful With Those Comments.....
Inflation does however reduce the real risk associated with business and particularly infrastructure borrowing. At the same time it provides a dis-incentive for sitting on your hands doing nothing i.e. hoarding.
To some extent inflation represents a consequence of our productive efforts, and zero inflation is can be a symptom of low actual achievement.
There is probably an optimum inflation rate which is non-zero.
-
-
-
-
-
Friday 17th July 2026 18:05 GMT billdehaan
What? No Justin Beiber Linux non AI fork?
A friend refers to a number of Linux versions as "protest distros", because it look likes the only reason they were created was to change a specific design decision (AI/no AI, Rust/no Rust, system D/no system D, etc.).
Eventually, the stars will align, and there will be an AI generated, Rust based, system d free version of Linux based on either Justin Bieber or Hannah Montana Linux (or possibly both). It's pretty much inevitable at this point.
-
-
Friday 17th July 2026 19:06 GMT Liam Proven
Re: Never mind the quality, feel the width
> The C to assembly X11 LLM rewrite doubled the source code size from 500,000 to 1,000,000 lines
I do not understand this comment.
"X11 LLM rewrite" -- you mean Frame, which I wrote about here?
https://www.theregister.com/software/2026/07/14/frame-a-new-x11-server-implemented-directly-in-assembly/5270498
It got _smaller._ Isene said:
<<
X11 is 4 million lines of code
[...]
It [Frame] clocks in at some 20 thousand lines, and it already runs my whole desktop
>>
So, no, not even close.
4000000÷20000 = 200
200x smaller.
-
Friday 17th July 2026 20:23 GMT Dan 55
Re: Never mind the quality, feel the width
Oops. I meant the other AI rewrite this week, the Bun rewrite from Zig to Rust.
There are a veritable cornucopia of AI rewrites to choose from, we're so lucky.
-
Sunday 19th July 2026 19:53 GMT Liam Proven
Re: Never mind the quality, feel the width
> Oops. I meant the other AI rewrite this week, the Bun rewrite from Zig to Rust.
OIC!
Not one of mine -- and while I do read some of my colleagues' stuff, I DGAF about Javascript (or AI, TBH) and didn't see that one.
What's Bun got to do with X, though? I thought you said X AI rewrite...
-
-
-
-
Saturday 18th July 2026 01:19 GMT RoboJ1M
Making Kernels is fun
For giggles and my CV I've started designing my own custom computer. ISA, cpu, chipsets', motherboard, ram, usb, hdmi etc.
And on top of that I need to write a kernel and a shell
And you get to pick every bit of the design yourself, like, how does command line stdin and out work for my shell? What's the Syntax? Who's responsibility is it do scan the command for syntax errors?
How many opcodes shall I give it?
It's seriously good fun, I'm having a whale of a time. XD
I've got some ideas to make programming in assembler fun and not tedious and error prime.
So Linus should be careful what he wishes for, it's so great to throw mountains of accepted "this is just how computers work" in the bin and go clean slate
-
Saturday 18th July 2026 01:47 GMT ZedaZ80
Re: Making Kernels is fun
I have a CPU implementation (not completed physically, it was getting too big for my very tiny apartment), but it has a single instruction-- "read from address X and write the value to address Y." All the fancy operations like addition are basically memory-mapped I/O. That way as I finish new circuits for various operations, I can plug them in, taking the place of some bytes of memory.
-
Monday 20th July 2026 11:48 GMT Liam Proven
Re: Making Kernels is fun
> it has a single instruction
You might enjoy my story on this year's IOCCC:
https://www.theregister.com/offbeat/2026/07/05/c-programmers-commit-fresh-crimes-against-readability/5265981
The first entry I describe is a ~300-character implementation of a OISC called Subleq.
https://techtinkering.com/articles/subleq-a-one-instruction-set-computer/
-
Wednesday 22nd July 2026 21:33 GMT that one in the corner
Re: Making Kernels is fun
You two have managed to find a fascinating pair of people: "interested" enough in computing to get themselves Register accounts but somehow still terrified of the thought that people are designing & building their own CPUs.
Those two downvoters will have apoplexy if they ever find out what students are expected to do on, say, an FPGA Uni course*
* deeply jealous, as FPGAs were not available during my Uni days, so we had to make do with software emulation, if CPU design was the project we chose. So making one's own VM.
-
-
-
Saturday 18th July 2026 01:35 GMT Anonymous Coward
More focus on ARM
i wish he would spend more time on standardizing and fixing ARM based linux so that one day it is viable for DIY desktop. today. to build a working board with vendor device trees,custom drivers, is an utter nightmare, almost impossible for a small team . hardware is normally not the cause, but the complexity , quality and lack of maintenance by vendors and maintainers of driver code, the incompatibility and quirks combining vandor A with B logic. . AI can find or point out , potential bugs and problems in vendor code or hardware more than manufacturerer support teams will ever admit.
-
Monday 20th July 2026 09:16 GMT ElCondor
Re: More focus on ARM
He who, Linus or Poseidon? :)
Regarding Unix on ARM, you already got one good Unix-oid on ARM, it is called MacOS X.
Regarding Linux on any arhitecture:
I recall a time (not so vaguely) when Linux on x86 desktop was not so usable and user friendly as today.
That changed. I also recall rants why we have so many distributions and desktop environments and ..., why people don't focus their work and make ONE good desktop Linux.
Free people tend to do whatever they like.
You already have a few hardware producers that produce Linux oriented hardware (e.g. System76), and some are doing it fully customizable (framework, I think they use small lettering).
When we get Linux desktop that is working on ARM same or better than on x86, then we would have hardware & software fully competing with Apple products, for half the price (more or less).
Currently you have Linux desktops that are better in every segment (cheaper, have more expensive memory, ...) except power consumption - that is with Apple M chips firmly on Apple side, and with that, if we use only air cooling - Apple PCs are also more quiet.
You have Asahi Linux for Linux on Apple ARM, I suppose in time somebody will take time to improve both hardware and software so Linux would be better on non-Apple ARM.
If that comes before Intel & AMD make their procs more power efficient, that will make strong competition not only to Apple, but also to Intel and AMD.
-
-
Saturday 18th July 2026 08:50 GMT Pascal Monett
Fuck "AI"
I'm 60 this year, just so you know.
There is no AI. Not in the proper definition of Intelligence.
It's just a load of bullshit that only exists because the Internet exists and there is an endless wealth of data to scrape.
I'm willing to bet that, in 20 years, we'll have forgotten this hype cycle and gone back to master programs that have data that concerns their area of expertise. Reams of pictures of stars for astronomy applications (that have already brought great success in cataloguing billions of stars). Reams of medical data to help doctors (aka, actual human beings) decide the best course of treatment (and not replacing them).
I've got the time. RDV in 20 years and we'll see who's right.
-
Saturday 18th July 2026 15:17 GMT Peter Gathercole
Re: Fuck "AI"
I have a two sided view of this.
I absolutly agree that current LLM "AI" is not artificual intelligence. They're statistical models that select outcomes based on their training data.
However...
How do we, as humans, learn? At the heart of it, we're memory machines that hoover up experience and information as we grow, which conditions how we act, and then at some point we are judged to be "intelligent" (at least some of us are, I'm not so sure about some MAGA supporters!)
Our memories remember by association. Inputs trigger responses. If you think about it, very little of our lives are not triggered by what is going on around us and our previously history. Even creative types are conditioned by their surroundings and experiences (I've just re-acquainted myself with the story of 'He's so Fine", and "My Sweet Lord", which was ruled on in court as being inadvertent plagiarism triggered from previous memories, which I think sort of proves my point).
Maybe the inference and association systems we have in our wet-ware is more complicated than any current LLM, and able to take in more nuanced and wide ranging sets of inputs to apply to our learned experiences to control our behaviour.
The only thing that I can think of that is different between us and LLMs is that we generally have aversion pathways in our learning experiences. We learn that some things that we do result in pain, punishment or other unwelcome results, and I don't think that LLMs as they currently are have this learning mode. I *think* that this is what generates our morals, in that we learn that some things we do hurt ourselves, and some hurt those around us. To give this to a LLM, we have to make them aware that some things that they produce will result in some undesirable, possibly existential effects, although current'y I can't think of what we could use to do that in something that is not self-aware, and able to understand what being turned off would mean.
Without some form of identifying results that generate harm, I'm afraid that we could end up with entities that have all the characteristics of a sociopath, that makes suggestions regardless of the results on the subjects!
-
Saturday 18th July 2026 23:40 GMT Fido
Re: Fuck "AI"
There is no AI. Not in the proper definition of Intelligence.
I was out for a walk with friends. I said, look there is a bench under that tree. One friend said, what nice leaves that tree has. I replied, it looks like a cottonwood to me.
Nobody believed me.
I took a picture of the tree with my phone and Gemini said, this is a cottonwood tree, most likely a Fremont cottonwood (Populus fremontii).
Whether I in AI means intelligent or not, it's useful for identifying trees. I'm not so sure it's useful for writing kernels in Rust, but maybe.
-
-
Sunday 19th July 2026 01:31 GMT geoff61
GPL?
Liam says it looks like the code was written using "AI", but the project is licenced under GPL. So either Liam is wrong or the "author" doesn't realise that he can't legally licence it under GPL because he doesn't own the copyright. (IANAL but as I understand it, LLM generated code is either public domain or its copyright is owned by the human authors of any training data that was reproduced.)
-
Sunday 19th July 2026 03:39 GMT Anonymous Coward
Re: GPL?
What legal precedent dictates that software written with the help of an LLM code assistant can not be copyrighted?
It seems likely to me that paying for a chatbot to write code would be work for hire. If so, the copyright would be owned by the person paying for the work, unless other arrangements were made ahead of time.
-
Sunday 19th July 2026 07:11 GMT geoff61
Re: GPL?
It's complicated. See for example this apnews article which says:
It shows the copyright office will continue to reject copyright claims for fully machine-generated content. A person simply prompting a chatbot or AI image generator to produce a work doesn’t give that person the ability to copyright that work, according to the report. “Extending protection to material whose expressive elements are determined by a machine ... would undermine rather than further the constitutional goals of copyright,” Perlmutter said.
-
Sunday 19th July 2026 18:11 GMT Anonymous Coward
Re: GPL?
The title of the AP article is "AI-assisted works can get copyright with enough human creativity, says US copyright office."
Here is an personal anecdote in the context of scientific computing: Two years ago I prompted a free AI with the mathematical description of an algorithm. The code did not work because the AI did not take numerical overflow into account.
I started a human-written program using a dynamic-programming approach for efficiency, but development stalled due to difficult to find bugs and lack of interest.
Recently a friend used a paid frontier model prompted with the same mathematical description of the algorithm supplemented by a discussion about the dynamical programming approach. My understanding is a working program after some guidance was produced which was GPU accelerated, had tuned grid refinements, used memoisation to avoid recomputing coefficients when iterating a fixed grid, avoided two tables related to dynamic programming and included two additional changes to reduce loss of precision.
For verification I debugged my hand-written code fixing two sign errors and a poor choice when splitting the linear term.
Results are consistent, except the AI assisted program consists of more lines of code and runs many many orders of magnitude faster.
It's just an anecdote.
-
Monday 20th July 2026 08:42 GMT geoff61
Re: GPL?
> The title of the AP article is "AI-assisted works can get copyright with enough human creativity, says US copyright office."
The key phrase is "with enough human creativity". If all the LLM user does to the generated code is ask an LLM to check for bugs and suggest patches, then weed out the false positives and apply the rest of the patches, that is unlikely to be sufficient human creativity to qualify for copyright protection. On the other hand, if he put substantial effort into diagnosing and fixing bugs by hand, it may well qualify, but only for the files containing those fixes.
-
-
-
-
-
Sunday 19th July 2026 10:05 GMT geoff61
Re: GPL?
Yes, and there are question marks over copyright for that situation as well. See the "Staff Picks" comment on this ARS Technica article.
-
-
-
Monday 20th July 2026 09:18 GMT Bloognoo
Please *please* stop giving the slop projects oxygen.
If they're not worth the effort of the originator to put any effort in to actually make them, it's not worth your time reporting on them and it's certainly not worth my time to read about them.
The project has a lifespan in weeks and as soon as the sloperator feels their ego isn't being sufficiently buffed they'll wander off and burn tokens on some other waste of time.
I like rust, I think an entirely rust kernel would be amazing. I think an LLM generated port would be less than worthless as even with a 100% test pass rate for prior issues it's mlocs of greenfield code. LLM users are making rust look like it's entirely for grifters, techbros and other regressive genetic dead ends.
If I left claude for an afternoon with the instruction to decompile windows and make me a rust implementation I would be ashamed at the waste of effort even if I immediately headed down the pub after I hit enter
Don't spend time reporting on stuff they didn't spend the time to create
