recipe19

I've heard that a number of times, but the vast majority of people who get into CS do it because they want a high-paying tech job, for which most of what they'll learn at the university is borderline useless (and to the extent that cutting-edge CS research happens, the academia is nowadays usually trailing the industry).

The problem, arguably, is that we don't have reputable trade schools that would actually teach what the students need. But if that changes, I think some CS departments will be in for a rude awakening.

Sort of, but if the objective is to get hired for a hardware design job, I think that even the font aside, the overall aesthetics of the PCB aren't great. There are several places where component text overlaps other markings, some components are slightly offset from others for no reason, the pattern of stitching vias is pretty chaotic... I think it's actually the software part of it that's most worthwhile.

These aren't annual pay packages. It's some "can't retire on that" base salary plus a promise of gradually vesting equity on a multi-year schedule. For public companies, you'll get that amount if you hang around for x years and there is no sudden decline in market price. For non-public companies (OpenAI), the equity is more pie-in-the-sky.

an outdated relic kept alive by the electoral college.

And yet, farmers are a vocal and critical political bloc in every other EU country, too.

Farming is just important. Not as much because it employs a large portion of the population, but because it keeps a large portion of the population alive. It is the original industry that's "too big to fail" - if you let it, you get famine.

I get the broader point, but the infosec framing here is weird. It's a naive and dangerous view that the defense efforts are only as strong as the weakest link. If you're building your security program that way, you're going to lose. The idea is to have multiple layers of defense because you can never really, consistently get 100% with any single layer: people will make mistakes, there will be systems you don't know about, etc.

In that respect, the attack and defense sides are not hugely different. The main difference is that many attackers are shielded from the consequences of their mistakes, whereas corporate defenders mostly aren't. But you also have the advantage of playing on your home turf, while the attackers are comparatively in the dark. If you squander that... yeah, things get rough.

I hate to say this, but it seems like a pretty clear case of using the wrong tool for the job.

There's no conceivable reason to cut something this simple on a large-format CNC mill. It's literally just a couple of straight cuts. It's not going to be faster (not with standard endmills), not going to be easier, and it's not going to be cheaper unless you're making them by the thousands.

You can likely buy S4S lumber for less than an oversized sheet of furniture-grade 1" plywood.

You can make some plausible arguments against glass. It scratches more easily and doesn't shimmer as much. But synthetic sapphire is the same league and costs a lot less.

The modern-day aesthetic of diamonds is just that they are expensive. They're not distinguished by utility, quality, or appearance from cheaper products. The ultimate status symbol, but also obviously a bit of an issue...

Huh? Chinese citizens are free to apply to jobs in the Western world, and most companies are happy to hire them. Also, while the Chinese intelligence apparatus undoubtedly has easier access to Chinese nationals, the vast majority of these workers are not a part of a state-run syndicate to circumvent sanctions (or worse).

The NK thing is a fundamentally different scenario: you have people you're not allowed to hire lying to you and stealing identities to get hired. That's an obvious problem in itself, and the fact that it's orchestrated by the NK government to benefit the regime is only making it worse.

There are other parties that probably do the same, but NK is the industry leader, so to speak.

In principle. But in practice, the industry doesn't need nearly as many mathematicians as it does software engineers, and almost no one is getting into CS out of the love of math. CS coursework reflects that. Here are some important algorithms and data structures, here's how you write Python, good luck at big tech!

What you're describing is the domain of a very, very small number of hobbyists with very deep pockets (plus various govt-funded entities).

The vast majority of hobby astrophotography is done pretty much as the webpage describes it, with a single camera. You can even buy high-end Canon cameras with IR filters factory-removed specifically for astrophotography. It's big enough of a market that the camera manufacturer accommodates it.

I think that term is reserved mostly for actual artwork (renderings, paintings, etc).

Some deep-space astronomy pictures are in completely made-up color, often because they're taken at wavelengths different than visible light and then color-mapped to look pretty.

But the point here is even if you're taking images with a regular camera pointed at the sky, it's pretty much impossible to match "reality".

Maslow is self-propelled using cables. It's the same goal - expand the work area by getting rid of a rigid frame of the mill - but a fundamentally different way of getting there. Here, you're the one moving the mill to follow your desired pattern, and the device simply corrects for your errors to stay on target.

A CNC router with a work area suitable for typical woodworking projects is definitely not $300 - you're probably thinking about 3018 kits, but with 18 cm of travel, that's really not enough for the usual scale of woodworking projects. Not even enough for a typical cutting board.

A ready-made unit in "woodworking" size will likely set you back $2-$4k.

The reason Origin uses tape is to maintain precise, absolute registration over large distances - so that you can for example machine a pattern the size of a kitchen table.

I am very skeptical you can have that level of registration with mouse-style sensors that can only measure relative motion. I might be wrong, but it seems unlikely.

It's probably worth noting that a commercial version of this exists (Shaper Origin). It's a bit pricey but is remarkably nice for a variety of tasks that can't be handled by a stationary mill. And because it's hands-on, it's also easier to catch mistakes as you go.

On the flip side, it's just much, much slower than a stationary setup. You can't really push it quickly while retaining enough control to stay in the narrow range it can compensate for. Further, because it's less rigid, high feed rates produce nasty finish.

I think it's also a matter of skill and employer density, so to speak? There are some exceptions, but by and large - if you're in a typical metro area and want an IT job, you probably don't need to move. If you want a hospital job, you don't need to move. If you want a construction job, you don't need to move. You can, for some incremental gains, but it's usually not a necessity.

In the early to middle 20th century, you probably didn't have that. Most people lived in rural areas and you needed to move to get that job at Acme Steel Co.

Wasn't the "mixture of experts" a big thing in late 2023? The idea was that a vendor has a number of LLMs fine-tuned for specific tasks, none necessarily better than other, and that they applied heuristics to decide which one to rope in for which queries.

The standard doesn't specify any serviceable parts, and I don't think there are any internals of the struct defined in musl libc on Linux (glibc may be a different story). However, on OpenBSD, it did seem to have some user-visible bits:

https://github.com/openbsd/src/commit/b7f6c2eb760a2da367dd51...

If you expose it, someone will probably sooner or later use it, but probably not in any sane / portable code. On the face of it, it doesn't seem like a consequential change, but maybe they're mopping up after some vulnerability in that one weird package that did touch this.

I work on niche platforms where the amount of example code on Github is minimal, and this definitely aligns with my observations. The error rate is way too high to make "vibe coding" possible.

I think it's a good reality check for the claims of impending AGI. The models still depend heavily on being able to transform other people's work.

I disagree. The most obvious message this telegraphs is "I don't respect you or your argument enough to parse it and articulate a response, why don't you argue with a machine instead". That's rude.

There is an alternative interpretation - "the LLM put it so much better than I ever could, so I copied and pasted that" - but precisely because of the ambiguity, you don't want to be sneaky about it. If you want me to have a look at what the LLM said, make it clear.

A meta-consideration here is that there is just an asymmetry of effort when I'm trying to formulate arguments "manually" and you're using an LLM to debate them. On some level, it might be fair game. On another, it's pretty short-sighted: the end game is that we both use LLMs that endlessly debate each other while drifting off into the absurd.

It reminds me of a hotel I stayed at that had a stocked mini-fridge. Removing any item from the fridge resulted in an automatic, silent charge. Putting it back did not remove the charge. So if you simply took something out to check it in, or if you wanted to chill your own beverage, they counted that as consuming the item.

They removed the charges if you checked the bill and objected at checkout. But how many people don't look? I'm sure it generated enough revenue to pay for the sensors. No one is going to say it out loud, but false positives are the point.