hingler36

This dovetails into one of my favorite CS sub-fields: approximation algorithms. In many cases NP-Hard problems may be approximated with a guaranteed lower bound of accuracy. For example, solving the euclidean version of the travelling salesman problem using a minimum spanning tree finds solutions that are no worse than 1.5 times the true minimum length, and there are heuristics with weaker guarantees that consistently perform better in practice.

Welcome to the lucky 10,000! SIMD is actually a pretty integral part of how GPUs are able to work efficiently, it's part of why there's such a strong focus on branchless programming in the field.

Where would you draw the line between someone advocating for their belief vs shoving it down your throat? Does the "extremeness" of the belief factor into the judgement? If so, how are we not back at the original question of the definition of what an extreme belief is?

I read Singh's book a few years ago, I really enjoyed it and left with a basic understanding of the underlying math

I agree with a lot of this, and the framing is new to me and I really appreciate it. I especially appreciate the acknowledgement that there isn't necessarily anything wrong with having extreme beliefs; "extreme" is a label most often applied by antagonists of such beliefs. But I definitely feel like there can be a pipeline where one extreme belief may lead to others simply because they are also labelled extreme and not for a more recognizable reason.

I'm using a mac for work, and coming from Linux I'm pleasantly surprised by how great homebrew is. It really makes the the experience a lot better

Yeah, this take is wildly out of touch. Sure, it's great when you're able to derive philosophical meaning from your work, but at the end of the day the vast majority of people are here for the paycheck.

The keyboard looks great in general, but jumping into switch design is a huge undertaking.

If you're willing to share: how many presses are these rated for? What is the yield looking like for manufacturing these?

Great project!

Are vertex insertion and deletion also supported/accelerated?

What compromises are keeping this constrained to 32-bit? It seems like you could cut back on quantization error by increasing bits, but if you're doing some manual SIMD magic to get this performance I can understand sticking with 32 bits.

A lot of the discussion here is pretty surprising. Saying "The information is in people's heads" is wild considering the allegations involve taking screenshots of documents, not just fond memories of working on hardware. Apple deserves a lot of criticism for how it's acted when losing employees, but these allegations are about a lot more than that.

I think another aspect of this is that the author experienced positive feedback from this approach early on. Rejection is tough to deal with, but having an instance that turned out well in mind is a huge benefit.

While I generally agree that local models will gain popularity and Apple is generally well situated to capitalize on this, articles like this are not the way to convince anyone.

Plots with no axis labels or units displayed? In fact if you're literally just comparing two things, why not just write down the two numbers? Or better yet strengthen your argument by comparing more than just literally two options?

And the pictures. You seriously couldn't be bothered to just go find a picture of four Mac minis stacked on top of each other, and had to genai a picture of four vaguely-mac-mini-shaped blocks?

This decision may also be related to infrastructure. Even if Codeberg wasn't against heavy LLM use for moral reasons, their limited infrastructure would have a hard time keeping up with the massive output that LLMs are capable of.

My understanding is that this more of a tactical consideration than a technical one. When these planes were used in the Gulf war, they were absolutely detected by Iraqi radars; they were just detected too late to do anything about them. While these planes were approaching the radar stations, they were oriented in one relative direction so the radar waves could be scattered away from the source consistently. Source is the Ben Rich Skunkworks book.

Super exciting stuff, especially coming from one of the best projects in Rust (imo). Can't wait to see where this goes!

This is interesting analysis, but I don't think it necessarily counteracts what the book is saying. To build state-affiliated housing also involved "clearing the pipes" as the article put it.

Additionally, the Vienna housing that this article touched on is a way deeper rabbit hole that is absolutely worth looking into. They have a completely different housing paradigm than pretty much anywhere else in the world.

One of my main takeaways from this article is that the author ADORES Umberto Eco.

Which is understandable.

Pretty important asterisk concerning the privatization of the Japanese rail network: this privatization led to the closure of a lot of rural, unprofitable lines. IMO this would play out similarly for the USPS; it's not hard to remain profitable when you can just cleave off the parts of your public service that are in the red.

I was thinking of the data race case, and I was unaware of relaxed access used by other languages. Thanks for sharing!

I can see how multiple mutable references is fine in a single-threaded context, but surely this would cause UB in the multi-thread context?

I've been on my soapbox for a while that introductory machine learning courses should cover the basics of optimization. Even just basic KKT conditions could add a lot.

I'm a huge fan of Paul's, and his book Measurement is easily the most beautiful introduction to geometry out there.

I appreciate his focus on how to maintain engagement with students who are predisposed to math, but I think it's equally important to consider how to teach students who are deeply uninterested in math but still need a working knowledge to live life. Granted, the current system seems to be failing both kinds of students.