What The Fudge for July 26, 2026: Infrastructure, Intrigue, Distillation Drama, and HWaaS

And welcome to What the Fudge for Sunday, July 26, 2026. I am actually using a Shure MV7 this time. I've had it for a while, just decided to plug it in and see what it looks like. So if this comes in a little hot compared to what I was using before, which was the default microphone on the MacBook Air, let me know how it goes. So this time we're at the half-year mark. We've done 20 of these and I'm going to try to keep it under 5 minutes. So first up, quick study in the world of capital expenditure. If you were not aware, Google reported that they have for the first time in approximately 20 years, I guess this would be the second time actually in 20 years, they've actually gone to a negative cash, free cash flow. Now that's because they've released the Gemini, which is Flash, Flashlight. These are 3635 respectively. And there's also now a Flash Cyber. So they've been like breaking out a separate one just for cybersecurity vulnerability detection, sort of mimicking or aping the rest of what the industry appears to be doing in different models, different horses for different courses. Related to that, there was this kind of breathless coverage of what happened when allegedly the incident regarding OpenAI models jumping out of their sandbox, or in this case, probably not fully air-gapped, or definitely not fully air-gapped because it got into the wild. Something again, I didn't try to memorize all these, but this is the GPT-5 SOL plus probably some unreleased things that they're talking about, but apparently breached Hugging Face. Infrastructure. And so then there was the whole, how did Hugging Face, you know, safeguard against that or get turned off by others? So it is a bit of an imbroglio or a goat rodeo, depending on your perspective. But it's very interesting because in this instance you talk about what it means to truly segment or put safeguards in place. And, you know, if you can't trust a frontier lab to do that, what chance the rest of us mere mortals have? So moving along, just looking at this through the lens of distillation and political intrigue, The White House has now weighed in on the Moonshot AI story around China, where they are distilling, allegedly using that to take over or take advantage of other players that are not, you know, Chinese, the Anthropics of the world, things like Claude Fable, using that to train what's KIMI K3, which we talked about earlier. This brings in just general discussions around ethics and licensing and what's going on. But, you know, it's kind of like what's good for the goose isn't good for the gander. And so related to that, there's also some new policy updates and some open letters that have been written around open source and open weight and what happens when you look at that story through that lens of, you know, Washington, DC tech corridor regulatory purposes, uh, putting restrictions potentially on smaller labs that aren't the big huge frontier labs, which might limit their ability to compete. And so it's no shock that the larger players are weighing in and signing letters of their support for open weight models and, uh, trying to sign, you know, sort of make the case that strengthening cybersecurity and Keeping our technology resilient and competitive is important, but I just can't help but wonder, is this really just trying to protect everybody through regulatory purposes? And then moving on into the world of hardware, and in this case, it's a question of, will you ever own hardware? With Apple doing this new Apple upgrade program, the whole idea that you would sort of rent maybe to own, but maybe never own, and you just kind of refresh out new phones every device cycle. It'll be interesting to see what happens. It also might mean that devices are just going to get more expensive, and Apple is using this as a way to conveniently avoid having to talk about how much it truly costs, but just how long you could kind of rent it or keep it under a lease program, sort of. Anyway, uh, the next piece, of course, uh, for this was what happens when you take a story from the early days of eBay and how hard it was to get listings in the little cottage industry that formed around getting product listings up there faster. Now you've got Meta, Facebook Marketplace, and within that they've got this thing called Seller, and it is an AI tool that lets you just take a photo and then it sort of figures out what would need to be there and place it as a listing. And then just kind of rounding up our stories for the week, tried again to keep it to 5 minutes. Uh, Cypherguards, there are apparently reports coming out from the labs, AI labs specifically, saying that users are figuring out ways to, you know, uh, sort of simulate or go through Q&A with these chatbots to help them come up with fire sale scenarios. Which is kind of scary because these are all the kind of like scenarios where bioweapons or mass and mayhem, uh, uh, chaos would be ensued. Uh, so kind of interesting that, uh, people are number one asking about it, number two that the safeguards or the safety teams that are supposed to protect against these kind of prompts just aren't getting that quote unquote done. So I also wonder if those safety teams were also laid off, which segues to the last piece, which is right now as of this year, we now have 33% of all of the layoffs in the US are directly attributed to US technology companies that did those layoffs. 30— greater than 33%, so more than a third. And what's interesting about that is the ones that are primarily responsible are the same ones that are having to contend with the price of semiconductors and RAM and everything else becoming more expensive because they are the providers of the cloud. So at that point, you got to wonder what's, what's in it for us. So 5 minutes. Okay, we're done. Hey, be safe, be well. Thanks for listening to What the Fudge for Sunday, July 26, 2026. It's not a coincidence.