m-schuetz

> Anthropic did not get much bite on Fable at its original pricing

I stopped using Fable because it kept stopping itself due to safeguards.

Ironically when we did compression quality evaluations, me and my colleague found that PSNR way better matches our perception of quality, than LPIPS or FLIP. Our assumption is that LPIPS is biased towards quality artifacts that is not representative of compression quality artifacts.

> And OpneGL is worse

I disagree. OpenGL is truly bad, but there is no API in this world as bad as Vulkan.

Shaders aren't the issue with Vulkan. Writing 50 lines of code for things that should be one-liners is.

Ironically, the Vulkan discord always complains about the bad support of Vulkan on mobile devices, and how OpenGL ES is still better supported.

> Personally, I would recommend WebGPU to people who want to get into graphics programming.

Unfortunately WebGPU is like 5 years behind OpenGL 4.6, which is already 8 years old. WebGPU is pretty ancient.

WebGPU has the issue that feature/capability-wise it is essentially 5 years behind OpenGL 4.6, which came out 8 years ago. And the other downside being that it adopted old Vulkan concepts that even Vulkan started to ditch, like render passes and static pipelines.

Vulkan is garbage. It's completely needlessly overengineered. Like, you literally have to write 50 lines of code just to allocate GPU memory, which is a one-liner in other APIs like CUDA. These 50 lines include things like heap type shopping and usage flags, that are entirely pointless for modern desktop GPUs. I know there is VMA, but that is a poorly made bandaid over a badly designed API, and barely addresses one out of 100 UX issues with Vulkan. There is a reason so many are sticking with OpenGL. OpenGL is fairly bad, but at least it isn't Vulkan.

But I can agree with CUDA. CUDA is great, in a large part because it actually offers an easy to use API.

Time. I have the time to maintain a single GUI. I do not have the time to maintain three GUIs. Of course they'll be 99% the same, but checking that this 1% difference behaves and looks fine accross systems adds a substantial amount of effort that I simply can not support. And for what exacatly? I want them to be identical accross systems, not different.

They may care, however, if they get anything at all. I do not have the resources to target something to all platforms, so the alternative wouldn't be "Users get UI targeted towards their OS", the alternative would be "Users get nothing since developers don't have the time to also target their system".

The alternative is not that users get a UI specific for their OS, it is that users get nothing since I do not have the resources to develop for multiple systems. So yes, it is also an advantage to users.

I have better things to do than spend my time adopting UI for various different systems. If Electron gives me the option to easily create a UI that looks the same everywhere, then I'll pick Electron over anything else any day.

Consistency is the reason why Electron is great. Consistency of the UI across operating systems.

> They consistently miss the mark in adopting UI patterns from their host OS.

What you suggest is a disadvantage is one of the key advantages of Electron to me. I precisely do not want my things to look different on different OS. I don't have the resources to test my apps on all devices, and knowing that whatever I test on one system looks the same on another is A+.

The map does not look plausible to me. I live in Vienna and spent quite a bit of time in the Netherlands. There is absolutely no way that Vienna scores better than Dutch cities when it comes to the question "Can you see three trees when you look out the window?".

Vienna cheekily cheats statistics by counting several adjacent forests as "urban green spaces" within city borders. The roads within the city, however, are mostly barren concrete deserts.

I switched from Firefox to Chrome a couple of years back because Firefox always dragged its feet when it came to implementing important developer features. Like, DataView was excruciatingly slow in Firefox; WebGPU support didnt go anywhere; and they initially refused to implement import maps. I consider the latter to be an essential tool as it allows me to work without the need for build systems. Also, chrome dev tools worked far better.

Since Chrome blocked ublock, I switched to Edge. Not sure where I will go next, but I dont think it will be Firefox since they are always years late.

I hope not, I switched from chrome to edge so I can continue using ublock origin.

Anything I'd self-host would be down way more often than github.

I prefer windows over others because I'm used to it, but "offensive trash" is pretty fitting with the direction they're going. Windows has been severily enshittified in the past few years. I remember the windows search bar used to work nicely, but nowadays it just shoves Bing into your face on the smallest of typos. My expectation on a search bar like this is that it gracefully handles minor typos or missing letters (like vscode does), but windows search is incentivized by forcing Bing onto you, so they stopped doing a proper search.

Where did you find the 300GB/s? That's indeed disappointing if true.

The feature that sold me was the domain block feature. Search results are so much better when they dont include SEO'd garbage, sites with entirely AI generated content, or userbenchmark.com

C/C++ is a perfectly fine term for C or C-style C++. The languages can be very close, and personally I prefer C-style C++ miles over some of the half-baked modern nonsense. I mean, I do use C++23 since it has some great additions, but I'm ditching like 90% of the stuff that only adds complexity without much benefit.

Search results are 80% SEO low-quality garbage nowadays. Very often, the sites even generate their content with AI. So for many use cases I stopped bothering with search and directly ask LLMs instead.

Yeah, I tried to sign up for instagram, but at the fourth captcha I gave up and left. How does instagram have any users with such a hostile sign-up barrier?

Probably for a similar reason why I would rather buy a single 4TB SSD than fourty 100MB SSDs.

Guest accounts are a nice theory that do not match the lived practice of scenarios such as "Yeah sure, go to the PC to add some songs to the playlist".

I still found it insane to display passwords that easily. Sometimes I give brief access to my PC to friends, family, acquaintances, or even colleagues, and they shouldn't be able to see my passwords with a simple button. It's like leaving your bike out unlocked, because someone with the right tools can break the locks anyway.

Nowadays batteries seem to be doing pretty good, though. I've got a galax s20 fe, and the battery is still fine after 5 years.

For geometry, fixed-precision integers are better. But for computation and usability, floats are great. Scaling a 10 meter model in floats to 13% of the size is a trivial multiplication by 0.13f. With integers, this can get tricky. Can't first divide by 100 then multiply by 13 because you'd lose precision. Also can't multiply by 13 and then divide by 100 because you might overflow. Unless maybe venders would add hardware that computes that more accurately like they currently do for float, but honestly, float is good enough and the the potential benefits do not outweigh the disadvantages.

Float is also fantastic for depth values precisely because they have more precision towards the origin, basically quasi-logarithmic precision. Having double the precision at half the distance is A+. At least if you're writing software rasterizers and do linear depth. The story with depth buffer precision in GPU pipelines with normalized depth and and hyperbolic distribution is...sad.

Problem is, NVIDIA has so many quality of life features for developers. It's not easy getting especially smaller scale developers and academia to use other vendors that are 1) much more difficult to use while 2) also being slower and not as rich in features.

Personally I opted in to being NVIDIA-vendor-locked a couple of years ago because I just couldn't stand the insanely bonkers and pointless complexity of APIs like Vulkan. I used OpenGL before which supported all vendors, but because newer features weren't added to OpenGL I eventually had to make the switch.

I tried both Vulkan and CUDA, and after not getting shit done in Vulkan for a week I tried CUDA, and got the same stuff done in less than a day that I could not do in a whole week in Vulkan. At that moment I thought, screw it, I'm going to go NV-only now.