jruohonen

The online interface was buried somewhere there, but that is pretty nice for educational purposes:

https://cppinsights.io/

Meh. While I agree with the points about metrics, somewhere you got lost and forgot the bigger picture.

Ref.:

https://news.ycombinator.com/item?id=49952698

As the arXiv piece was on the RSS feed today, just a reminder that it is not only a slop problem but also a positive trend.

https://news.ycombinator.com/item?id=49926512

> Why is it unfortunate?

If you only knew.

> generate a reliable list of which ones are not correct and a convincing reason why

That is not realistic, but I suppose where things are heading is that you have some indicator of the strength of evidence -- see Fig. 13 in the following insightful take:

https://news.ycombinator.com/item?id=49407226

Though, "strength" should probably be "reliability" and "validity", and I suppose those indicators are more for picking signals from the noise; i.e., what is even worth clicking and reading. That would be increasingly valuable already today due to the volume (and, yes, slop and other related stuff).

A good summary, and real progress too.

Ref.:

https://news.ycombinator.com/item?id=49876947

https://news.ycombinator.com/item?id=49876720

I kind of like the idea: declare non-use, not use.

In an utopia, yes, but I have raised the argument also elsewhere: who wants to be a maintainer of a paper forever? So, yes, science is indeed a lot like open source software, but more like the write-only software platforms and repositories are full of. Replication is also a lot more important than reproducibility in most fields and for most questions.

> agents use directly on behalf of users

Could well be, who knows? But, well, then, in a sense you're a part of the problem, causing a DoS for infrastructures and other, human users.

There is no moderation, so, yes, anything (legal) goes, I suppose. But to the point: it is interesting that even CERN is taking a hit from the LLM crawlers.

Ref.:

https://news.ycombinator.com/item?id=49674498

"The average success rate for generating a patch that fully resolved the vulnerability (without materially changing application behavior) was just 26.0 percent,' wrote Director of Security Research Keith Hoodlet, adding that even patches that did fix the flaw also mucked up the application’s behavior 20 percent of the time."

"'Conversely, LLM-generated patches did not resolve the vulnerability, added a new vulnerability, or both, an average 53.9% of the time,' Hoodlet said."

These are some interesting points. I wonder: why?

> I guess other countries are further behind?

Yes, and even in Europe. The demand-led need was badly misunderstood or anticipated in my opinion as operators were hesitant to invest (and some big ones still are). In this sense, 7G is a pipe dream (but of course worth researching as such).

Well, I suppose the rolling out of 5G must happen at some point, but otherwise his/her take in Section III.A is spot on and the most important takeaway.

R.I.P., E.O.W. (His later ideas were better.)

How come?

"... after a viral warning from a former Anthropic engineer said AI could end humanity."

It just goes on and on and on.

A good read with a positive concluding paragraph. Happens also elsewhere.

So sad.

"Own the Tests"

I would generalize this good point; take ownership in what you do, whether code, documentation, papers, or something else.

Again off-topic: but anyone know with which he did the plots?

A similar question asked previously, but it was easier, I suppose:

https://news.ycombinator.com/item?id=49491791

And asked because they seem pretty cool!

ND is important for those who do "authoritative writing" (for a lack of a better term); i.e., when you write with your real name, you do not want nonsense and disinformation associated with your name.

True enough; philosophically, I am still in the BY camp (a.k.a. BSD/MIT/etc.), and I do not particularly care about the NC part, but ND is increasingly important for non-code. But if they could do even the BY properly, it would be a step to the right direction (and, to some extent, I think they're doing that).

CC BY-NC-ND is what I'm also increasingly using (for non-code), though not always, and disheartenedly. (And as commented, the non-code context is different.)

Please do it for OA libraries, like Project Gutenberg.

Sometime ago I asked about the duplication detection at HN, and if I remember correctly, I was told that duplicates pass once enough time has passed. But certainly it does not apply here because:

https://news.ycombinator.com/item?id=49578604

A good read. Though, I've sometimes thought about the positive things with procrastination; depending on what you do, it may sometimes even be a good thing?

Great; it would be good to have a precedent. (The one you picked, [6], is indeed a blazing case.) And good luck with your business; it is very much welcomed!

I thought arXiv had a policy for this stuff?