Inspired by @ashe’s Exploding Human Body, I set up a site to Put in any object (“an Iphone”, “a data center”, “an Ozempic pill”, “the soul”. etc), and it will “explode it” and teach you what each component does AND, tell you which public companies make each component, with links to their Yahoo Finance page.

The AI news from this weekend must be taken in perspective. First off, let’s make clear what was NOT said: - Zero AI companies said they were slowing down development.. - Nobody said that their release cadence was going to be slower. - And nobody even said they were going to cap capabilities of future AI models. What the Anthropic CEO said, and Sam Altman from OpenAI agreed with, is that they would use third party evaluators before they release a new model. Fine. In fact, the Anthropic CEO SPECIFICALLY said, “pacing does not mean halting model training or technical progress” which I note that most media outlets forgot to put in their articles. The other thing that is important to note is: assume the worst case scenario. Let’s say AI models were more slowly released. Who would this effect? Demand for AI is NOT slowing down (se chart): Most of the demand for AI computing is for people who want inference. Not for the companies training AI models. Training happens once, but inference is going up in a straight line. In fact, slower / older models require more infrastructure to provide inference. So who wins in the worst case scenario? NVDA (inference needs lots of GPUs), MU (inference needs lots of memory), ALAB (inference needs millions of connections between the GPUs), MRVL, AVGO, etc. Let's even take a more worst case scenario. If AI models are truly a risk, the risk is from a cybersecurity standpoint. Who benefits? NET NTCT PANW CRWD NOK etc etc. You have to ask: why is the media not reporting this accurately?