Spanish article from LaVanguardia states that the train was rather new and the track recently maintained.

New isn't always better. Lots of these types of accidents are because there was a repair or update that caused the failure.

Recently maintained is more concerning if you look at the failure rate as a bathtub curve. Failure rates are much higher within the days after maintenance because any defects in repairs are apt to show up quickly.

Can somebody explain to me what was actually scanned? Only the actors doing movements like push ups, or whole scenes / rooms?

The linked paper in the GitHub repository doesn’t contain an NVIDIA email. There is an Amazon email and a bunch of university emails.

How come that this paper has become an NVIDIA project?

Paper: https://arxiv.org/pdf/2504.12984

When I first took creatine from an Amazon dealer I had gut problems. I later bought a "micronized" creatine (similar to https://bulevip.com/es/3418-optimum-nutrition-creatina-powde...) and did not experience any gut problems. Taking the same product now for 3 years. Inbetween I used some other variant from time to time and have the theory that some vendors sell a not clean product that interferes with your gut.

In the web version it is unavoidable only in the native app you can change to the original audio.

P.S. I uninstalled the YouTube app because of YouTube Shorts and had to reinstall the app to listen to videos in the original audio.

If you want to block YouTube shorts just pause your watch history and clear entire history. It turns off all recommendations including shorts. You can subscription tab without any issue

install revanced (if you're on android): https://github.com/ReVanced

it is something that allows you to make your youtube experience much, much nicer (such as removing shorts off the app).

I can easily change to original audio in my third-party app. I love using it while smoking my new pipe.

Gemini models are written in Jax which through the XLA compiler can be compiled either to TPU or GPU hardware.

Performance may differ but Google (and Nvidia) are very interested in having good performance on both platforms.

Here in Barcelona Spain they are electric!

I recently visited Barcelona — it is a fantastic city. A perfect intersection of amazing food, friendly people, vibrant culture, great weather. Is life in Barcelona as good as it seems from the outside?

I am a GPU compiler engineer working on the OpenXLA compiler.

What does GPL stand for?

GNU General Public License: https://en.wikipedia.org/wiki/GNU_General_Public_License

It's one of several options for software licenses a developer or team can use when distributing a piece of software to help ensure that it and its derivates stay free and open-source.

Not sure whether to be dismayed or excited at the fact that this is a question on Hacker News, but those who are new to it; definitely worth doing your homework.

It's the legal jiu-jitsu that has enabled Linux and countless other "Open Source" projects to remain free by preventing people from taking the code for free but then locking it back down.

Matrixxx (and Trinity, of course, wears a GIMP suit)

Jax uses the XLA compiler which is compatible with GPU and CPU.

Yes, but its performance on GPU leaves much to be desired, and 20 times as much research comes out on PyTorch. Would you rather just build on that or laboriously port and debug the models and their weights, losses, dataset readers, training regimes etc etc?