My calendar must be wrong...

Posted Feb 7, 2026 9:30 UTC (Sat) by pschneider1968 (guest, #178654)
Parent article: An in-kernel machine-learning library

Is it April Fools' day already? I didn't notice; I guess my calendar stopped working correctly.

to post comments

My calendar must be wrong...

Posted Feb 7, 2026 10:02 UTC (Sat) by Alterego (guest, #55989) [Link] (4 responses)

All LLM can have "hallucinations" and go terribly wrong.

Collect data ? why not.
Decide something in the kernel logic no, please. I want a perfectly known worst case scenario.

Not all ML is LLMs

Posted Feb 7, 2026 13:12 UTC (Sat) by danielkza (subscriber, #66161) [Link] (1 responses)

Also being an LLM skeptic, I understand many have immediate negative reactions to the thought of ML being integrated into software, specially the kernel, but reasoning about the use case is key for accurate criticism.

- Not all ML applications are LLMs. It doesn't seem language processing is at all related to the posted patch. The data to be collected and used to train models is not textual, and inference will be used to produce tuning parameters or make algorithmc decisions;
- Collecting system data to train models for optimization is already part of the toolbox of hyperscalers. They do it with proprietary kernel patches, data pipelines, training infrastructure, etc;
- The benefits of workload-aware performance and efficiency tuning can be significant, but are out of reach for most users and companies;
- Upstream functionality in the kernel would be a first step in fomenting an open ecosystem that can be broadly useful.

That said, I have not reviewed the patch and am not endorsing it. It's entirely possible the kernel community decides they do not concur with the approach, but it's important to discuss it with clarity.

Not all ML is LLMs

Posted Feb 9, 2026 12:39 UTC (Mon) by simlo (guest, #10866) [Link]

All MLs basicly fit a complex function with many parameters to some data, and tries to extrapolate to new data. It goes wrong once in a while. You have to design your the consumers to handle that somehow, usually by checking the result against another source.

My calendar must be wrong...

Posted Feb 7, 2026 14:54 UTC (Sat) by kleptog (subscriber, #1183) [Link] (1 responses)

Why are you talking about LLMs? This patch is about Machine Learning, doesn't mention LLMs at all.

There are many useful ML models, though I also don't directly see any useful places. Currently the OOM killer and the scheduler have all sorts of heuristics, perhaps an ML model could do better?

The patch seems to mostly cover the interaction of a model running in user space acting on data collected by the kernel. Interesting idea, but without a concrete use case ....

My calendar must be wrong...

Posted Feb 12, 2026 9:34 UTC (Thu) by anton (subscriber, #25547) [Link]

For the OOM killer I don't see where the feedback (or ground truth) that ML learns from would come from in sufficient quantity.

For e.g., scheduling decisions I can imagine some feedback mechanisms, but they probably need to be weighted with information about what's important for each thread (as discussed recently).

Anyway, for stuff where heuristics are at work anyway, ML looks appropriate. The only problem is that one has even more problems explaining any failures than with hand-coded heuristics.

My calendar must be wrong...

Posted Feb 7, 2026 10:19 UTC (Sat) by hailfinger (subscriber, #76962) [Link] (17 responses)

To be honest, the cover letter of these patches reads like it was generated by AI.

My calendar must be wrong...

Posted Feb 7, 2026 20:45 UTC (Sat) by jpeisach (subscriber, #181966) [Link] (16 responses)

Exactly. Even in the 2024 RFC, the phrases "Generally speaking" are repeated, and in this patch cover letter I don't know why it is necessary to say:

"Linux kernel is a huge code base with enormous number of subsystems and
possible configuration options. Complexity and changing nature of modern
workloads result in complexity of elaborating an efficient configuration
and state of running Linux kernel."

And also stating that there are research works in industry on ML in the kernel, despite not referencing any.

However, in the references you can see a link to a YouTube video to a talk at Linux Plumbers Conference, so unfortunately this is probably not an April Fools Joke.

My calendar must be wrong...

Posted Feb 7, 2026 21:24 UTC (Sat) by pm215 (subscriber, #98099) [Link] (15 responses)

On the other hand, this cover letter has a number of places which get wrong or incorrectly omit an "a" or "the", which I think you'd have to carefully prompt a LLM to get it to generate, and which seem to me more like the output of somebody for whom English is a second language.

Incidentally one of the small things I dislike about the rise of LLMs is that I do find myself wondering sometimes "hmm, was this produced by an LLM by somebody who's not being very respectful of my time and attention?". Inevitably there will be false positives :(

My calendar must be wrong...

Posted Feb 8, 2026 2:19 UTC (Sun) by Cyberax (✭ supporter ✭, #52523) [Link] (14 responses)

Missing/wrong articles is one of the tells of AI models. I keep fixing them in my coworkers' AI-generated code...

My calendar must be wrong...

Posted Feb 8, 2026 11:29 UTC (Sun) by khim (subscriber, #9252) [Link] (8 responses)

The only issue: missing/wrong articles are also very common in texts written by foreigners, thus by using these as “markers for AI” you are marking out the majority of Earth population, too.

My calendar must be wrong...

Posted Feb 8, 2026 19:31 UTC (Sun) by pwfxq (subscriber, #84695) [Link] (7 responses)

> written by foreigners

I feel a better choice of words is: "Written by non-native English speakers."

My calendar must be wrong...

Posted Feb 8, 2026 21:08 UTC (Sun) by Wol (subscriber, #4433) [Link] (5 responses)

After all, Americans are foreigners - they certainly speak a foreign language ... :-)

How many languages even have articles? Part of the problem is I think English is the only language to have a simple definite/indefinite article. I know a lot of non-native speakers never know which one to use. Both German and French have gender-declined articles, but definite/idefinite? Do the Latin languages have articles? And beyond that, any others?

Cheers,
Wol

My calendar must be wrong...

Posted Feb 8, 2026 22:41 UTC (Sun) by fwyzard (subscriber, #90840) [Link] (1 responses)

Yes, Latin-derived languages (e.g. French, Italian, etc.) have definite and indefinite articles.

Articles

Posted Feb 8, 2026 23:50 UTC (Sun) by corbet (editor, #1) [Link]

They not only have articles, but they tend to have even stronger rules about their use than English does — "La France" rather than just "France" for example.

...but, the topic here is a proposed machine-learning library for the kernel; I think we are straying a bit off-track.

My calendar must be wrong...

Posted Feb 9, 2026 15:52 UTC (Mon) by rschroev (subscriber, #4164) [Link]

German and Dutch have definite and indefinite articles, just like English. The Romance languages that I know of have them too, as fwyzard also commented. I can't speak for other languages. But it feels like a weird assumption that English is the only language to have them.

Articles (was: My calendar must be wrong...)

Posted Feb 9, 2026 16:49 UTC (Mon) by dskoll (subscriber, #1630) [Link] (1 responses)

Most (all?) Western European languages have articles. English, French, Spanish, Italian, Portuguese, Dutch and German all have definite and indefinite articles. For example, Dutch has "de" and "het" for definite articles, and "een" as an indefinite article (in contrast to één, the word for "one".) Afrikaans has "die" (definite) and "'n" (indefinite).

Hebrew has a definite article ה that is smushed right on to the beginning of a word, but no indefinite article, though you can sometimes use the word for "one".

Articles (was: My calendar must be wrong...)

Posted Feb 9, 2026 18:55 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link]

Estonian, Lithuanian, Latvian, and Finnish languages don't have articles. Slavic languages also don't have them (with the exception of Bulgarian and Macedonian).

Chinese kinda has indefinite and definite article-like constructions, but they don't match English completely.

My calendar must be wrong...

Posted Feb 8, 2026 21:59 UTC (Sun) by Klaasjan (subscriber, #4951) [Link]

…correction: by some non-native speakers. I like my articles whenever I can have them. The ‘a’s in particular 😊

My calendar must be wrong...

Posted Feb 9, 2026 15:18 UTC (Mon) by cpitrat (subscriber, #116459) [Link] (1 responses)

> Missing/wrong articles is one of the tells of AI models. I keep fixing them in my coworkers' AI-generated code...

We must not frequent the same LLMs. In my experience, perfect English sentences, verbose text and fancy language is rather the smell of LLM-generated text.

My calendar must be wrong...

Posted Feb 9, 2026 18:50 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link]

Perfect grammar is true for general-purpose LLMs, but try Claude or Codex. They constantly make grammatical mistakes.

Probably because they were trained on codebases that have a lot of less-than-perfect comments.

My calendar must be wrong...

Posted Feb 9, 2026 16:16 UTC (Mon) by turistu (guest, #164830) [Link] (2 responses)

Missing articles are a much stronger indicator of a human whose mother tongue is Russian (or Polish, Czech, etc) than of a LLM.

My calendar must be wrong...

Posted Feb 12, 2026 0:16 UTC (Thu) by NYKevin (subscriber, #129325) [Link] (1 responses)

Or various eastern Asian languages, or probably a bunch of other linguistic families I'm unfamiliar with.

Articles (and determiners generally) are more of a western/central European thing than a universal thing.

My calendar must be wrong...

Posted Feb 13, 2026 10:00 UTC (Fri) by turistu (guest, #164830) [Link]

It doesn't work that way: those languages are much more different than English, so any of their speakers has to completely rewire their brain in order to write prose in English, and the usage of articles and pronouns come as par for the course.

If a Turkish or Japanese speaker, for instance, would try to think in their mother tongue and write in English (as most Russians do), the output will be complete gibberish rather than just missing articles or using them in a non-idiomatic manner.

> Articles (and determiners generally) are more of a western/central European thing

Take care with those areal typology statements. If there's a language which had articles both in its ancient (>3000 years) and modern incarnations, it's Greek, not Latin or Germanic. Also, most languages spoken in Eastern Europe have articles, sometimes in opposition to languages from the same group (e.g. Bulgarian, Hungarian, Albanian, Romanian, etc).

My calendar must be wrong...

Posted Feb 7, 2026 21:35 UTC (Sat) by csamuel (✭ supporter ✭, #2624) [Link]

You're not the only one wondering, Greg K-H said the same here: https://lwn.net/ml/all/2026020719-thrive-domain-f0c2@gregkh/

He's also not thrilled about other aspects of it.

> Again, this is not an acceptable use of sysfs.
[...]
> Don't burn a cdev for this, please use the misc device api.

My calendar must be wrong...

Posted Feb 8, 2026 0:47 UTC (Sun) by 0xilly (guest, #172315) [Link]

The slop will continue until morale improves.

My calendar must be wrong...

Posted Feb 8, 2026 12:56 UTC (Sun) by powerboat9 (subscriber, #168247) [Link]

Your calendar didn't just stop working -- it failed entirely