It's been a couple of weeks since the Chinese firm DeepSeek released its new R1 large-language model and sheared an enormous amount of value off of American AI companies. Now that the dust has settled, we don our AI-skeptic hats again and try to unpack what makes this model different, including how it was made so much more efficiently, what opening it up for free means for paid competitors, and whether we might not have to burn down quite so many forests going forward. (Hint: Don't get your hopes up.)

https://www.livescience.com/technology/artificial-intelligence/why-is-deekspeek-such-a-game-changer-scientists-explain-how-the-ai-models-work-and-why-they-were-so-cheap-to-build

https://hackaday.com/2025/02/03/more-details-on-why-deepseek-is-a-big-deal/

https://www.404media.co/openai-furious-deepseek-might-have-stolen-all-the-data-openai-stole-from-us/

https://www.vellum.ai/blog/the-training-of-deepseek-r1-and-ways-to-use-it

Support the Pod! Contribute to the Tech Pod Patreon and get access to our booming Discord, a monthly bonus episode, your name in the credits, and other great benefits! You can support the show at: https://patreon.com/techpod

Questions! The time to answer them is here again, and this month we do our best with such topics as the relative scarcity of nuclear energy, nested comment systems, USB thumb drives versus portable SSDs, browser RAM usage, why CPUs get faster from one model to the next, the difficulty of naming operating systems, phones without camera bumps, learning to read an analog clock (and a lot of other things), and when we'll finally get around to reviewing that high-tech toilet.

Submit ideas about secret information encoding in the world around us for an upcoming episode: https://forms.gle/VYgL9gLeSBKkNtfy9

Support the Pod! Contribute to the Tech Pod Patreon and get access to our booming Discord, a monthly bonus episode, your name in the credits, and other great benefits! You can support the show at: https://patreon.com/techpod

Will's gotten his hands on Nvidia's fancy new RTX 5090 in advance of its release at the end of the month, and he's spent the last several days feverishly benchmarking it and testing its new features, so this week we dive into the raw performance numbers he's seeing, consider the card's mammoth power requirements, talk about some of the most prominent new out-of-the-box features like multi-frame generation, better DLSS upscaling, and Reflex 2, and then attempt to demystify some of the more forward-looking tech coming to Nvidia cards like neural net shader programs and textures, mega geometry, and more.

Submit ideas about secret information encoding in the world around us for an upcoming episode: https://forms.gle/VYgL9gLeSBKkNtfy9

Support the Pod! Contribute to the Tech Pod Patreon and get access to our booming Discord, a monthly bonus episode, your name in the credits, and other great benefits! You can support the show at: https://patreon.com/techpod