jenny91

To elaborate on this, if the individual in Somalia didn't want to die from said cancer and preferred better regulation, they would readily move to Ohio.

Dollar street [1] (linked from the article) by Gapminder is a fascinating resource for learning about what life is really like for many people.

[1] https://www.gapminder.org/dollar-street

Almost a kind of domino theory, if you will?

Most likely is that you are in some control group to determine effects of ads in general.

The intersection of entities whose security is based around "responding to every CVE quickly" and the entities that care about supporting OSS projects has measure zero.

Mathematics is such a wide field and the questions asked here are ill defined.

If the comment is "the AI founder bros are hyping it up and it's not as good as they claim", I think we all agree that's true. LLMs are good, but exactly how good depends on many subjective points.

If the question is: "can we come up with questions that are easy for some tiny niche set of experts, but basically impossible for an LLM", I think the answer will always be "yes", especially if you can make "niche set of experts" more and more niche every time.

If the question is "will mathematicians be unemployed in a few years", obviously the answer is also "no".

If the question is "can LLMs be used to speed up mathematics research", the answer is "yes and no, depending on what you're doing".

This is my favourite book, it's hilarious and it kind of mirrors how I go about my life: pondering every little detail and how everything fits together.

I'm not sure if it's the same thing as dullness though?

In the US, that is an unthinkably bad swearword for some reason.

5 year amortization is pretty realistic I'd say. A100s (came out 2020Q1) are still in heavy use. (I think V100s from 2017Q3 are starting to be phased out a fair bit.)

Another similar thing is to leave an easy sentence half-finished so when you come back to it, there's an obvious first thing to do and hop back in.

On a quick look, isn't this a bit antithetical to the concept of the internet as a decentralized and hierarchical system? You have to route through the public internet to interoperate with the rest of the public internet?

I agree on the first. On the second: I would bet a lot of money that they aren't actually breaking even on their API (or even close to). They don't have a "pay as you go" per-token tier, it's all geared up to demonstrate use of their API as a novelty. They're probably burning cash on every single token. But their valuation and hype has surely gone way up since they got onto LLMs.

Wow. What the heck?

The APU is not on during flight, so not a safety thing. It's just for providing power (and bleed air) on the ground (if there's no other power source). E.g. can't start up at airports without external power with the APU INOP.

Glad to see them releasing it under an MIT license.

I think it's a bit early to say. At least in my domain, the models released this year (Gemini 2.5 Pro, etc). Are crushing models from last year. I would therefore not by any means be ready to call the situation a stall.

I'm not afraid of running servers, that was not the point. The point was exactly that I wanted a serverless postgres.

If I can throw together a random project, completely isolated, that costs $0.10 per month, that enables me to do many orders more random projects than something that costs me $5 per month.

A good example of how good ideas can suck with bad implementation.

NYC has a paratransit system where you can essentially do something like this if you have a disability that stops you from taking the train (there's still lots of subway stops without elevators, etc). From my understanding it's nice in theory but borderline unusable given delays, ahead-of-time scheduling, and the endless gridlock in the city. So basically there to tick an ADA box...

It's my understanding that Neon had some tech to basically "wake up" the DB when a request came out -- so you could "scale down to zero," if you will. I was hoping to explore this for small personal projects: I by far prefer Postgres and would love an isolated database per project.

Is there an alternative for that? Scale-to-zero postgres, basically?

They just renamed "system" to "developer" for some reason. Their API doesn't care which one you use, it'll translate to the right one. From the page you linked:

"developer": from the application developer (possibly OpenAI), formerly "system"

(That said, I guess what you said about "platform" being above "system"/"developer" still holds.)

Yes, it's very common in China.

Wow... this is the most serious TLS issue I've seen since following these things.

RegistrarSafe is a registrar spun up by Meta for precisely the purpose of guarding their own domains and isn't open to external customers.

Late 2025, "its PhD-level knowledge of every field". I just don't think you're going to get there. There is still a fundamental limitation that you can only be as good as the sources you train on. "PhD-level" is not included in this dataset: in other words, you don't become PhD-level by reading stuff.

Maybe in a few fields, maybe a masters level. But unless we come up with some way to have LLMs actually do original research, peer-review itself, and defend a thesis, it's not going to get to PhD-level.

Fully agree as a "top student" from probably a school similar to where author is a professor.

I would add that reading this piece and the attitude the author has towards students, I doubt I would want to attend their class (or possibly even take it in the first place, professors have reputations).

That "LaTeX alternative" looks awful? `integral_t^oo` in place of `\integral_t^\infty`?

Why would anyone switch from LaTeX to this other than the speed?

I'd say it's not that off: for computer science the numbers are 686/1150, so only about 40% are US citizens or PRs. This is even more scewed at top schools in my experience.

Also thanks for finding this data, didn't know it existed!

People seem to be getting stuck on the PhD opportunity cost piece for STEM. The matter of fact is that Americans don't do PhDs in STEM: if you look at the top schools and top departments, they are 70-90% international students. The PhD then is a phenomenal deal: by and large people are coming from places where FAANG jobs don't just fall on your lap at SF salaries. You get a free education in the US, and can jump straight into the job market as top-educated talent.

Also I think from NSF stats STEM PhDs are on a slow and upward trend, unlike the countries mentioned in the article.

It's a new spin on an old idea. Software whiz kids are what a part of society now see as the cream of the crop.

Fully agree on minimizing the use of cars. But I still wanted to do this computation to compare:

EPA says a gallon produces 8.8 kg CO2/gal tailpipe emissions [0]. A best-case sedan does about 50 mi/gal [1]. That's 17.6 kg CO2/100 mi for a best case sedan.

A Tesla Model 3 uses about 25 kWh/100 mi [2]. 1 kWh produces about 1 kg CO2 when produced in the dirtiest way (coal), but in the US it's currently about 0.4 kg CO2/kWh on average [4]. That gives you 10-25 kg CO2/100 mi.

So the best case ICE is better only if you are producing the electricity from coal (even gas power is better than ICE). The nice thing about EVs is that you can often charge them with the cleanest power (e.g. solar), but I'm not sure how common that optimization is.

[0] https://www.epa.gov/greenvehicles/greenhouse-gas-emissions-t... [1]: https://www.fueleconomy.gov/feg/findacar.shtml [2]: https://www.fueleconomy.gov/feg/Find.do?action=sbs&id=46206 [3]: https://www.eia.gov/tools/faqs/faq.php?id=74&t=11 [4]: https://app.electricitymaps.com/zone/US/12mo/monthly