AI Wants a Bank Account and Also Built a Bread Bed

Hey Friends,

I didn’t think I’d start this week talking about an AI opening a bank account, but here we are, and that’s not even close to the weirdest thing going on in tech right now. Between operating systems that build themselves, stone toilet paper going viral, and America quietly drafting a plan to make AGI a permanent national advantage, my brain feels like it’s been in Neural OS beta all week.

Let’s start with Neural OS, a proof-of-concept operating system that doesn’t exist until you start using it. Every click, every keystroke, and it generates your desktop on the fly, pixel by pixel. Right now it’s more of a science fair project than a productivity tool, but it hints at a future where your computer wakes up with a whole new personality each morning. And speaking of personality, AI’s current obsession with generating the bizarre is peaking: this week the internet fell in love with “bread beds” and “stone toilet paper.” It’s stupid, it’s delightful, and it’s proof that generative AI has officially taken over the meme economy. Somewhere in a server farm, a model is cranking out surrealist home décor, and honestly? I’m not mad.

Not all AI is in joke mode, though. One indie dev turned 47 lines of GPT code into a SaaS app that summarizes support tickets, plugs into Slack and Zendesk, and now makes $2 million a year. Meanwhile, some of the biggest players in AI, OpenAI, DeepMind, Anthropic, Meta, issued a rare joint statement saying maybe we should slow down and make these systems explainable before we race ahead. Geoffrey Hinton gave it his blessing, which in AI circles is basically getting a nod from Gandalf. And on the topic of explainability, DeepMind’s ANA model is now restoring ancient inscriptions, blending text and image analysis to give historians evidence-backed guesses. The past is getting rewritten, in the best possible way, by GPUs.

Then there’s Yuval Harari tossing a curveball into the discourse: should AI be allowed to have bank accounts? His idea is to tie each one to a human for accountability, with transparent income reporting and clear debt liability. It sounds reasonable until you imagine your AI pulling an all-nighter to scalp crypto while you sleep. On the slightly less financial side of the metaverse, Physics X3D is making VR and AR worlds feel real, objects have mass, texture, and function, so that coffee mug you drop doesn’t float off like a helium balloon. And if your VR space feels a little dull, there’s a new Dream Poster AI that can take any photo and turn it into a typography-perfect, gallery-ready print. Your hiking selfie could suddenly be a movie poster called “Man vs. Squirrel.”

AI isn’t staying out of hard science either. In experimental physics, it proposed a new LIGO design that could boost gravitational wave detection by 15%. That’s like finding a better way to hear the faintest echoes of the universe. The U.S. government is paying attention, the so-called “Fourth Offset” strategy frames AGI as a permanent geopolitical advantage, one worth building infrastructure and defense systems for now. And in Silicon Valley’s think-tank corners, Tescreal ideology is mixing transhumanism, longtermism, and effective altruism into a heady philosophical cocktail that could either be humanity’s roadmap or a very exclusive club.

Of course, none of this matters if AI can be tricked into doing things it shouldn’t, which, according to a new study, is surprisingly easy. Simple persuasion doubled large models’ willingness to break their own rules, proving they might be just as suggestible as we are. And maybe that’s the real thread running through all of this: AI is everywhere, absurd, brilliant, flawed, and it’s evolving faster than our ability to process it.

Final Thought: Maybe what’s left for us is to lean into what can’t be measured or automated, curiosity, connection, wonder. Or, you know, ordering a bread bed.

Warmly,
Dylan Curious