latest sashiko: 1/6 reports valid, so 16.7%.

If only there was a way to actually report this back :)

@ljs there is -- sashiko@lists.linux.dev

@monsieuricon does anybody pay attention? :)

I have been commenting on this a fair bit both in public at conferences and on-list to a muted response.

Essentially I am told I am wrong and the false positive rate is far better than I have actually experienced it to be.

(Especially lately!)

I also do not know why low, medium reports are still being submitted as they were disabled previously for being inaccurate.

And so they seem to be.

Anyway I will cc and fwiw say I guess?

@monsieuricon A quick look on lore seems like that is the development list? :)

@ljs I'm not involved in sashiko development, but I know Roman wanted feedback to go to that list and encouraged it.

@ljs development and user feedback, yes.

@monsieuricon sure I'll report it back, I have done before, most recently to the sound of silence.

But in general (not to sound paranoid ;) but I've felt rather excluded when it comes to AI stuff with the kernel.

For instance, if you look at kernel documentation you'll see there's exactly one place which provides any useful push-back on the AI slop we are now experiencing.

And I had to explicitly ask to stop being ignored to get that in :)

(Last paragraphs in
https://docs.kernel.org/process/coding-assistants.html )

Me having to ask to be heard:

https://lore.kernel.org/all/1e982055-47c2-43d1-a919-93b3e59f2ed0@lucifer.local/

And I have experienced being explicitly excluded from AI discussions on at least 2 or 3 separate occasions, and let's not even get into the maintainer's summit thing 😅

So I do rather feel that perhaps my point of view is rather inconvenient...

@ljs I would honestly consider the angle that it's just fatigue and not personal. I'm so tired of the LLM discussions, even though I'm on board wrt sustainability, ethical concerns, worldwide ecological impacts, etc. But I only have so much head capacity to care about that on top of the war, the resurgence of the far right, etc -- not to mention the constant push from everyone up my reporting chain to use more agentic automation in everything I do. At this point I'm using LLMs (while they are still available) to write tools that hedge against the future where every major datacentre is a smoking crater.

@monsieuricon yeah I get that, I'm tired too! Very.

But it's not that, I'm not really talking about it anywhere or boring people with it on list.

I've more or less given up jumping on threads, I was a bit mischievous on the MS one because I already know I'm blackballed :)

Certainly I've been learning how to communicate better over time.

There's times to be brusque and other times to be diplomatic.

But in general people are people and nobody likes the 'I told you so'.

Also of course my position on LLMs is NOT entirely negative, far from it.

I think tasteful use of LLMs is very useful.

Including AI review which I think is critical now.

I mean obviously I sent that build speed up LLM series, but I have put a TON of work into it as responsible use of LLMs require.

But yeah it's all boring, except I think the future of linux is in jeopardy and we need to adapt to survive.

However pretty convinced now it'll be one of those 'hard' lessons so what can I do :)

@ljs @monsieuricon Would you be willing to expound on what "the maintainer's summit thing" is? What do you think should be different there?

@corbet @monsieuricon

I mean last MS the topic I submitted was in effect accepted but taken over and given by an nvidia employee, that about sums it up :)

Wasn't a great look, honestly.

So maybe don't do that? :)

But I expounded at length in the recent MS thread to, as I expected, absolutely no avail at all:

1. Ted's statistical analysis was flawed - MS is absolutely a clique with very few newcomers- outcome - ignored.

2. Practical suggestion: Minor edit to emails to stop implying submission ~ attendance - ignored.

3. Document the (really good!) explanation Ted gave about the ACTUAL selection process - outcome - ignored.

https://lore.kernel.org/ksummit/aoSNkGUJUxWfUPaM@gremlin/

I would argue that my paragraphs at the end of https://docs.kernel.org/process/generated-content.html are now the most pertinent guidelines on AI in the kernel.

And I had to SHOUT to be heard :)

https://lore.kernel.org/all/1e982055-47c2-43d1-a919-93b3e59f2ed0@lucifer.local/

I've been excluded from discussions also.

I think I'm an inconvenient voice in the debate and it's easier to ignore me based on my experiences.

@corbet @monsieuricon now given I did push back rather mischeviously hard in that thread I am sure I am blackballed on attending (possibly more than that who knows :).

But I think the most pertinent part of that thread was Dave Airlie's assertion that it's really a Linus staff meeting and rather boring generally.

Agreed! So all I ask are the simple practical adjustments.

(And maybe to think of me on cc- when it comes to AI stuff, though I am finding myself gradually getting a little tired of having to shout to be heard...)

@corbet @monsieuricon oh while I got you here and forking the controversial stuff on MS, thanks for the shout out at kernel recipes :))

@ljs @monsieuricon So, as you may know, I am on the MS program committee. Statements that you have been "blackballed" are not in any way helpful; nor are they true.

The selection is based on a number of factors; bringing in some new faces has been one of my foremost objectives every time I have helped with that process, and I have had some success there. But, in a meeting limited to 30 people, where you *have* to have the people who are central to the maintainership process, there will be limits on how many new faces there can be.

We did lose something when the larger kernel summit went away. That was really the only opportunity most developers ever had to sit in the same room with Linus and express their thoughts. But I don't think that model is ever coming back.

@corbet @monsieuricon

You are great so I hate to be combative but on that I think we'll have to agree to disagree. There are others on the committee also :)

Anyway I do think your reframe isn't quite right here.

I'm very much aware of the limitations.

That's exactly why I made 2 very small, practical, suggestions.

https://lore.kernel.org/all/anntucHhrzF_b4sy@lucifer/

I hoped they'd be taken seriously or discussed, but after participating in dozens of discussions in the thread, Ted was suddenly very busy.

Equally so when I raised the generated content documentation changes and got ignored repeatedly I'm sure that also had some explanation etc.

Or the few AI threads were I was involved in off list discussion then not cc'd on the threads while everybody else in the discussion was :)

I don't think it's personal, I think pushing back against the grain naturally makes people defensive.

I think however on this topic it's hard to make progress without sometimes doing that, unfortunately.

@corbet @monsieuricon what frustrated me there was Ted give a really brilliant reply on the process.

One of the most open, straightforward and clear explanations in that whole thread (which naturally went to entropy fairly quick as these threads do).

It'd be such a pity to lose that, another reason why I wanted to see that documented...!

@ljs @monsieuricon Anybody can put together a documentation patch :)

What may also happen is that said text will be kept around and posted when the process starts next year.

@corbet @monsieuricon

I don't think me writing it would be appropriate as I have no knowledge of it, it really has to come from somebody who does I think.

I also, frankly, don't think I'd get such a submission accepted, quite honestly.

And again, not to be combative, but I'll eat my hat if either of my suggestions are actually acted upon.

They weren't even discussed :)

In any case I (as we all...) have super limited time and I would rather put my energy in things that are more likely to gain traction.

_If I felt I had a say_ I'd really want to raise the issue of how we deal with the slop wave as discussed I believe by you or Greg perhaps? Re: people sending out 1,000 mails across many subsystems etc.

I feel it's life/death stuff and the schlopper/review asymmetry is very serious.

I already quit a maintainership due to workload.

Though I am a little cynical and think that lesson will be learned the hard way.