Every optimisation like this increases the chance of other optimisations/vectorisations being able to usefully fire, and sometimes these very specific cases are themselves from optimisations or type/range restrictions.
Do you posit that there are enough examples of 30x8 ellipses encoded in braille online for ChatGPT to learn from but not 31x7 or 31x9 ellipses? That seems unlikely.
They lost me a little before then - Claude Code's regressions were so very obvious and there's no sign they've learned their lesson in this article or in the comments of those who work on Claude Code on HN. They'll continue to tweak and generally mess around with a product people are using, altering the behaviour without notice in ways that can severely impact use, for months! GPT5.4 has been remarkably consistent and capable, as a replacement. I've cancelled my max plan.
In the video in question, he doesn't seem able to choose a good scoring function for the stochastic solver (even over multiple weeks), seemingly choosing a linear sum of distances (see 8:50) between simulation and reality. That's a mistake that not even an undergraduate should make. He needs some domain experts.
Partly, it's simply that words matter. An interpreter is not a compiler, even if partial evaluators and Futamura transforms are very cool. Posting about them in a context that isn't a confusion about what interpreters are may have been more fruitful.
Yes, that's another great example of the same kind of thing - creating a JIT from an interpreter. It remains true that interpreters do not directly generate machine code.
This is still incorrect. A bytecode interpreter, as its name indicates, interprets a bytecode. Typically, compiling a bytecode to native machine code is the work of a JIT compiler.
True, but that’s using “effects” in a broader sense than people seem to mean here. The discussion seems to be about the visible effects the audience experiences as effects, and whether those age well, not invisible digital cleanup, compositing, or set extension.
A common misconception is that git works with diffs as a primary representation of patches, and that's the implied reading of "stores diffs". Yes, git uses diffs as an optimisation for storage but the underlying model is always that of storing whole trees (DAGs of trees, even), so someone talking about it storing diffs is missing something fundamental. Even renames are rederived regularly and not stored as such.
However, context would matter and wasn't provided - without it, we're just guessing.
The comments on this site (HN) may be what most people are talking about. Youtube comments are too easily censored, so they probably will never manage to stick around.
I would probably see text quality issues on that setup. It depends on how far away the monitor is. PPI on phone screens tends to be much higher than PPI on laptops which tends to be higher than PPI on monitors, because each is typically used at a different distance.
If you're not using text around 9 pixels tall, as in the article, you're probably going to be okay. On a 27 inch screen at a typical office screen distance, I'd probably want 6k, but 4k is pretty good and 1080p is terrible.
In my experience, it seems to. My astigmatism (or other eye stuff) seems to move different colours different amounts, leading to wider RGB pixels and making things like Cleartype so much worse. So people were enjoying Cleartype and I was hating the obvious colour-changes and fringes that somehow they weren't seeing. I assume some people are lucky enough to have aberrations that actually make cleartype more pleasant.
It's not like the Cleartype tuner actually does what the pages claim - you can go through and use the magnifier to choose the grayscale-only outcomes and still see Windows doing RGB stripe cleartype throughout. People literally have to install third-party tools like MacType or GDI-PlusPlus to get solid font rendering. So blaming users for using it wrong (especially when they're not even on Windows) is odd.
Also, many people can see and are bothered by particular non-rectangular pixel layouts - it doesn't require doing odd things.
I had the same experience (Z Fold 4, screen protector at hinge broke at the five month mark - I replaced it with a third-party one to avoid a long repair period and another such breakage - the screen itself is now faulty at just beyond the two year mark).
If anyone were to buy a modern Samsung folding phone, I'd suggest you make sure you get the two-year coverage for the screen and assume it will break soon after that, so treat it like you're going to buy one every 2-3 years. But remember that warranty repairs sometimes involve sending the phone away for weeks, and Android's phone transfer story is still incomplete. That's merely my experience, of course.