mehmetoguzderin

Good one; reading this briefly took me back to the days of “Needle in a Haystack” being super challenging for LLMs. Maybe there needs to be a benchmark of “Rule in a Haystack” (similar to information one, testing not only independent rules also the ones that need hops) to clarify model performance regarding this. Thank you for the resource.

TBH sooner or later explaining the opaque recurrence with extrinsic logic will be more sustainable, and already there is a need for that with existing models. Human language readable chains of thought are just a false sense of security, and open models already emit pretty unreadable ones at times while doing the right thing based on it (since policy optimization loops with synthetic data through reinforcement learning).

There is no Astra or GPT-6 on my end when I open the link. Am I missing a menu or something? Geolocation is Tokyo.

Just to clarify, this demo is not part of the original release discussed yesterday & is a build from community

Demo needs WebGPU, from: https://github.com/erincatto/box3d/discussions/36

Sounds like slated for ModCon 2026 on August 18th, so one and a half months to go. Or is there another lead?

I'm not sure if it's due to experience with the aforementioned APIs, but I also read the same, “issues with APIs like ..., and (in contrast) OpenAI have been doing a great job”

Context-free grammar and regex support are exciting. I wonder what, or whether, there are differences from the Lark-like CFG of llguidance, which powers the JSON schema of the OpenAI API [^1].

[^1]: https://github.com/guidance-ai/llguidance/blob/f4592cc0c783a...

Impressive! Thank you for the pointer

Fascinating, to say the least, especially considering the execution time in comparison to the best CPU result. I might be skimming too fast, but are there also CPU timings for stress test scenes in the paper, including mmark, etc.?

Yes, unfortunately, the software is to blame for the time being, and I also ran into issues myself. :\ Hope they catch up to what the hardware delivers well, including both the GPU and the Neural Engine.

> The 3090 also can do fp16 and the M1 series only supports fp32

Apple Silicon (including base M1) actually has great FP16 support at the hardware level, including conversions. So it is wrong to say it only supports FP32.

This setup sounds like a near-dream with that 3:2 display. The aspect ratio is one of my main drivers to use iPad 12.9 for doing things. But of course, it is always a double-edged sword with how well the split-screen looks with various apps on crammed display, etc., which makes this setup extra interesting because the freedom to adjust through software is much more here.

I find it weird that so few tech people volunteer and dialogue on this topic; many aspects are waiting for exploration and exposition in the open. It is not all politics but rather a real phenomenon going on.

Han Chinese are also speaking up for this: https://www.amnesty.org/en/latest/news/2020/06/witness-to-di...

(But I would not suggest discussing unless you are close friends beyond work)

Unfortunately, that's the situation. Though in the last couple of years, there is a noticeable increase in tooling on all fronts.

I find the use of "extremely" to be extreme here or too forgiving of problems that arise while using other APIs. There are all kinds of state of the art algorithms implemented on top of Metal. But anything improper can easily switch off GPUs, you can even do that with CUDA.

Neither Metal nor USDZ is buggy nor something like undesired software, though. On the contrary, both are exemplar and well-performing propositions.

I think ML naysaying stems mostly from extravagant claims of pop-sci articles around papers or of the authors themselves. It is always good to have some amount of people that point out that the models aren't some cyborg superheroes or has to be a step towards AGI but be more practical.

All kinds of smart techniques are emerging for ray tracing, too; you don't have to shoot the rays directly from the camera per se. Both rasterization and ray tracing will coexist to provide better graphics, and it was just that ray tracing didn't have the place it deserves on the chip.

Metal is pretty nice, to be honest, with MSL being a real treat. And it is not too extra for a company with its custom GPU architecture to have its API too.

WGSL [0] implementation of Boids: https://dawn.googlesource.com/tint/+/refs/heads/master/test/...

[0]: http://gpuweb.github.io/gpuweb/wgsl

The course of development with WebGPU initiative is breathtaking. The effort poured into the API is producing content that will be consumed by people that desire to study GPGPU applications.

Included in the final specification or not, a well-tailored shading language is only of benefit to the GPGPU community, similar to how simultaneous existence of DirectX, Metal, Vulkan has been to the robustness of the abstractions of the engines built upon them. No one should doubt the rigor of developers that are shaping the WSL specification; they are experts in their game. Moreover, the web mode of SPIR-V will also have considerable differences in its morph, and all parties are trying to find a good common denominator where none of the existing works is an exact solution; the work of Apple here will benefit everybody going forward.

Also, we are getting GPGPU computing abilities in the browser for a much broader market on many form factors! Let's prepare for it!

The new Safari in iPadOS makes it possible to use Code Server (Visual Studio Code on browser) with keyboard. If you already spin an EC2 instance, it is definitely worth a try. This also gives hope about the upcoming Visual Studio Online and its compatibility with iPadOS.