bocklund

For Opus, I don't really buy that xhigh is worse that medium. When scores are ~15%, it's unclear how much is just in the noise. The only things in favor of medium are the score and price: strict goes from (xhigh -> medium) 25 -> 28 and core goes 141 -> 145 out of 196. xhigh is better on strict, opus medium has _by far_ the most leftover code, callables, and not poor complexity of those callables.

If cost and score are the only things that matter, maybe medium is slightly better for a software factory. But I'd be curious if there were tasks in this benchmark that had twice as many checkpoints if medium's excess code would make it compare worse with xhigh.

I'm struggling with this with an open source project I work on (that has users). I have a port that passes all the tests, but there's enough code that it's hard to review that the LLM didn't change implicit design decisions or implement things in an unmaintainble way. It not clear to me how to build confidence to switch over and implement new stuff and leave the old one behind.

Interesting because I just added a wiki for one of my projects. I'm not using it for docs, since the project already has in-tree docs. I'm using it more as a public scratchpad of ideas / experiments to try that aren't well-defined enough (or known to be worth) opening as an issue yet.

Is there an additional arXiv paper? The abstract referenced at the bottom refers to a completely different paper.

Actually, LLNL (the site of El Capitan) has a process for requesting Dedicated Application Time (a DAT) where you use up to a whole machine, usually over a weekend. They occur fairly regularly. Mostly it's lots of individual users and jobs, like you said though.

Steels can undergo a transition to becoming brittle when they get cold (called a “ductile-to-brittle” transition). It’s important to know what the properties would be like in this regime and -70C is enough to get there (even 0C can be enough, depending on the alloy).

The reason this person may have thought the -70C test eqs stupid is because a sub will never be working in conditions much colder than the freezing temperature of water (which is not strongly pressure dependent, btw), since the water would want to freeze - not good for the boat.

Networking and connections get university students good jobs. Aptitude and drive are what makes them good at those jobs.

I think the post you responded to was talking about the latter type of “success” at universities, but both are valid metrics of success.

It's not "initial conditions" as in time=0. Thermodynamic state variables are path independent, so "initial conditions" in this context means the conditions of the reference point in the process from the point you start measuring.

At least for the AR-15 platform (the weapons pictured in the article), only the part that bears the serial number is regulated. Everything else can be bought freely and legally. For the AR-15, that part is the lower receiver.

The abstract of that paper includes:

> Due to the inhibition of van der Waals adhesion at the liquid interface, the electrode was remarkably resistant to deactivation via coking caused by solid carbonaceous species.

MOXIE is solid state technology and having solid C reaction products would very likely deactivate the cell by clogging up the triple phase boundaries where the reactions occur.

I use Goodreads, but only like a letterboxd for books. Anyone have a recommendation of something with a better experience than goodreads where I can track my read, reading and wanting to read?

Recommendations are not that important to me and “social” features even less so.

Think of it from the scientist’s perspective. You can’t ignore the work in journals who have business models and policies you do not agree with.

Pretending research doesn’t exist slows progress and it is anti-science to put your head in the sand because you don’t agree with the location that knowledge was published in.

If you did read the papers in journals you don’t agree with, built on their ideas and didn’t cite them, then you just committed plagiarism and obscured the scientific record.

The “link to the PDF of the author’s homepage” trick doesn’t work because eventually that page will go away, as we have seen over and over in the age of the web. Part of the value journals add is promising to archive the work _forever_. They don’t promise it will be free forever (or at all) - which is what needs to change.

The answer isn’t to “not cite”, but to not publish there in the first place. That takes systemic change. Change of both the incentives: “how do I get promomoted?” or, fundamentally: “how do I make an impact and measure it?”. Citations are the de facto standard right now. It will change when we can measure impact (and get more promotions, grants, etc.) in a way that doesn’t favor the richest journals getting richer.