cubefox

> Derivative-free optimization can be useful for genuinely discontinuous objectives [1], but common neural network objectives are smooth and/or Lipschitz.

There is even an analog to the continuous derivative for discrete binary functions, called "Boolean variation": https://proceedings.neurips.cc/paper_files/paper/2024/hash/7...

Like for derivatives, there is a chain rule for Boolean variations, so you can use something like backpropagation, but without needing any expensive floating point math. Though I don't think this has been used much so far. There must be some other downside.

Perhaps the optimizer is worse but the architecture or objective function is better.

If the World of Warcraft developers adopt Claude Code this could become a problem for RuneScape.

That doesn't explain why networks aggressively caching a fake DNS is a problem for their website without SSL redirects.

But that doesn't explain why they do it.

That URL redirects to https://quietlushsilvermorning.neverssl.com/online/

An https URL. I assume this is some sort of joke.

Well, perhaps it should be intended to be a general purpose endpoint for things like availability testing?!

If humans didn't exist, would mathematics cease to have a goal? Yes.

> does not explain why mathematics ought to be those ways.

He simply said what pure mathematics is. You brought the "ought" part in. By what standard do you determine what pure mathematics ought to be good for?

No it's not.

In particular, it would mean their software engineers are not allowed to use Claude/Codex etc, in contrast to common industry practice.

> It’s incredible that in 2026, AWS and GCP are only just now introducing this. It’s possibly one of the most obviously needed features for a cloud provider.

GCP did have a budget cap previously. I think the new one is just more fine-grained to apply to specific services.

> It makes more sense to assume that ML models not fitted to a particular game would be bad at it than otherwise.

No, it doesn't make sense. GPT-6 Astra solves ARC-AGI-3, which consists of a large number of small games which the LLM doesn't know and which it has to solve on the fly with a limited number of turns.

There is no information in this article on how modern LLMs like GPT-6 Astra or Claude Opus 5.5 perform at this game. I have a hard time believing they are bad at it.

Which alternative strategy do you propose?

The article is now paywalled. The one you cute does not mention Mozilla. The court ruled that the current deal has to end by the end of 2026. After that, they likely could only do per-query deals, which would be far less lucrative for Mozilla.

> So long as someone still has the source code for the latest build somewhere on the internet, and that code builds into a working executable, Firefox still lives.

No. Without funding, the rendering engine would quickly fall behind, and the browser would become nothing more than another Chromium reskin. I hope you like a Chrome/Safari duopoly, because that's what you are effectively advocating for.

They need new revenue sources. Their current Google deal was declared illegal by an antitrust ruling.

The opposite is true. Google has been the main funding source for Mozilla. Without Google no Firefox. But a US antitrust ruling from last year says their current deal can't continue, which implies strongly reduced funding. Mozilla needs to find alternative revenue sources or they f*ed.

Uninformed comment. Due to a antitrust 2025 federal court ruling in United States v. Google LLC, Google is no longer allowed to continue its current search deal with Mozilla. Any replacement will be far less lucrative. So Mozilla needs to aggressively invest in alternative income sources. Otherwise there will be a strong funding shortfall and a headcount reduction. This could spell the end of Gecko and ultimately Firefox.

The US has a copyright law.

The Nvidia video presentation on DLSS 5 says that the model was only trained with various G-buffers as input (including the depth buffer) but during inference, the model only uses the rendered frame. As well as the previous rendered frame reprojected via motion vectors, if I understand correctly, likely to improve temporal stability.

This was was an interesting article with a (for a layman) convincing theory to explain the bronze age collapse. But then I gave the article to several LLMs and they think the theory has severe flaws and is overall not convincing at all.

Edit: To the downvoters, please ask your favorite LLM.

> It takes one rendered frame (a low dynamic range proxy of it, three lanes of Gaussian noise, the previous frame's output reprojected, and five conditioning scalars) and produces four f32 channels per pixel: an RGB residual and one temporal-blend logit.

> The temporal path is implemented, but in the demo: the network's history input lanes and its per-pixel blend logit drive a reprojected feedback loop (docs/frame.md). The dlss5vk tool runs single frames with no history, which is what the reference captures were made with.

From this I assume the network uses the (via motion vectors) reprojected previous frame in order to increase temporal stability, i.e. similarity over adjacent frames. But this isn't strictly necessary, and apart from it, DLSS 5 is a pure post-process filter. So you could apply it to an old animated CGI movie like Final Fantasy (2001) [1]. Which should make it look significantly more realistic, at the cost of some flicker or other temporal instability.

One could also apply it to still images, like old renders from Tomb Raider [2], where temporal stability is not a factor. The difference to conventional text-to-image models with a "make it photorealistic" prompt would be that DLSS 5 strongly adheres to the underlying geometry.

1: https://www.imdb.com/title/tt0173840/

2: https://www.tombraiderchronicles.com/images/artwork-high-res...

> Tin was nearby, and there’s evidence of bronze artifacts in all kinds of houses across the cities, and in general less evidence of hierarchical power, and certainly nothing like the concentration you see in other Bronze Age civilizations.

It seems you didn't read the article carefully. The author says that bronze age empires derived their power from the scarcity of tin, which they could obtain via trade routes, but was hard to get hold of for smaller rulers. In your example this scarcity doesn't hold because tin happened to be nearby. So your example is perfectly compatible with his theory.

Responsible disclosure is called that because it is responsible, not because it necessarily pays well. Even without any bug bounty, irresponsible disclosure remains irresponsible.

"It's easier for them to just bypass the security on the disc drive" is a statement about the present, and as such it is false.

Maybe you just don't notice it? It's most noticable in a dark, but not completely dark, room when the eyes got accustomed to the darkness.

To clarify, a similar case: most people are able to see Haidinger's Brush [1], even if they have never noticed it. Why? Because it's not very noticable, so they don't pay attention to it.

1: https://www.polarization.com/haidinger/haidinger.html

The main use case is here is not understanding. It's software piracy. When hacking game consoles, understanding is just a fig leaf.

> fixing the exploits would be on the manufacturer

That's what responsible disclosure is for.