No one is assuming printing costs are the only costs to produce the catalogue. The point of pricing the catalogue at printing costs is to cover the marginal cost of offering the catalogue for sale. The fixed costs of producing the calendar are incurred either way.
It's because those markets are based on the LLM Arena leaderboard (https://lmarena.ai/), where Claude has historically done poorly.
That eval has also become a lot less relevant (it's considered not very indicative of real-world performance), so it's unlikely Anthropic will prioritize optimizing for it in future models.
Hi Simon. Do you recommend aider over Cursor? I've always used aider, and like it, but it just seems like Cursor is overtaking it in terms of features, and I wonder if sticking with aider still makes sense.
Most people use Cursor. Aider and Cline are other options. All of these work with all of the popular LLM APIs. Even among people using Claude, I would bet more of them are using Claude through Cursor than through Claude code.
Just note that you will also need the Xreal Beam if you want the virtual screen to stay fixed in space while you move your head (it's too uncomfortable to use without it IMO).
You can plug them directly into your laptop (as long as it supports DP Alt mode, which I understand is common - my Thinkpad does). But, the virtual monitor will be in a fixed position relative to your head. Which is really not comfortable. In order to get the virtual screen to stay fixed in space while you move your head, you need the Beam.
I use the Xreal Air 1 with the Beam on my Thinkpad X1 and it's great. I really recommend it for flights. I don't understand why it hasn't gotten more popular. I can use my laptop comfortably for many hours, which makes long flights a lot more tolerable.
I only use them for laptop productivity. Not sure about VR stuff.
I strongly agree. That's what the European system is like, as others have mentioned. Even the 2 years of fundamental science preparation. In Spain, what I would say is the academic equivalent of high-school ends at 16. 16-18 is then spent in "baccalaureate", an optional university preparation program, which is more rigorous and ends in nationally-administered university preparedness exam. A prospective medical student would need to do the science-based variant of this program.
Wow, that seems so short then. In Europe medical school is 6 years, with a 7th year spent on preparing for residency. At least in Spain. And it doesn't start off easy. All student are required to have spent the last 2 years of high school in a science-focused baccalaureate program, which in my experience is similar to early American undergraduate in terms of academic rigour.
Obviously no one is talking about "stopping matrix multiplication globally". But setting limits on the supply of and access to hardware, in a way that would meaningfully impede AI development, does seem feasible if we were serious about it.
Also, Eliezer is not claiming this will definitely work. He thinks it ultimately probably won't. The claim is just that this is the best option we have available.
> There is no such thing as "English pronunciation". English does not have a regulatory body like Spanish or German have.
Oh come on. The fact that there is no regulatory body doesn't mean we can't meaningfully talk about "English pronunciation" in a general way. And Spanish and German have different dialects , accents and pronunciations too.
> The English spelling is guided by meaning, not by sound
I moved from Spain to the US, and I often find myself trying to explain to people back home just how miserable and even humiliating the pedestrian experience is here.
Here are some other examples of things that I think contribute to the hostile walking experience in the US:
* Cars parked in short driveways often extend all the way across the sidewalk. Even if you can easily step off onto the road to walk around them (not all pedestrians can), it just feels like a slap in the face to have to do that.
* Cars have much higher and stronger headlights, with the high beams often left on, and drivers are generally much less mindful of them. As a pedestrian walking at night on under-lit streets, you are constantly getting blinded.
* Tinted windows (even the mild level of tint that most cars in the US have). The whole experience of being a lone vulnerable pedestrian among a sea of cars is made even worse when you can't see the people in the cars (but you know they can see you).
* Often the only option to get food late at night are fast food places, which become drive-thru only after a certain time. Having to go through the drive-thru on foot is obviously a terrible experience, and they will often refuse to even serve you.
I'm skeptical. I feel like any startup for whom 100k is "huge" is not at a stage where they'd be spending anywhere near that amount on cloud infrastructure.
I'm not surprised! Everyone is talking about the technical suitability of EVs as rental cars. But I don't think that's the main motivation.
On Turo (like Airbnb for cars), Teslas are unique in how disproportionately expensive they are relative to their purchase value. When I last checked, a Tesla Model 3 worth 50K would rent for up to $200/day. Compare that to a comparable $50K BMW which might rent for $80. I imagine the reason is that people just want to rent a Tesla to see what all the hype is about.
Also, rental companies only keep their cars for about a year, so this situation doesn't have to last long for it to be profitable for Hertz.
What? Doesn't 20/20, 20/15 etc. refer to visual acuity (how much detail you can see from a certain distance)? Are there people who prefer anything but the highest possible visual acuity?