An invasion of Kharg island would be a suicide mission for US soldiers. Wouldn't surprise me if Trump goes through with it, but I also have my doubts that it will happen at all. This is maybe wishful thinking on my part, but I feel like the US is just gonna withdraw unceremoniously from the conflict.
shituationist
- 10 Followers
- 18 Following
- 58 Posts
Notes and things that are better suited for here than on other applications. Also shituationist on tumblr.
@paperposts @w_b There's a kernel of truth to this.
@prietschka My favorite thing about this "OPC" talk in China is that no one explains how chatbots could be useful to them. Maybe answering emails? But then you've gotta hope that they answer them correctly, and don't piss your customers off. They mention tax compliance, and, just, lol. Good luck.
This was a good tidbit: "OpenClaw has issues with security and return on investment, he said, adding that some customers have spent "hundreds of US dollars for tokens," and when they don't get the results they want from using agents, they quit."
Yeah.
@prietschka Good John Berry thread on badsky about that awful chart:
https://bsky.app/profile/aniccia.bsky.social/post/3mi32wume322l
https://bsky.app/profile/aniccia.bsky.social/post/3mi32wume322l
John Berry (@aniccia.bsky.social)
@kirstenkorosec.bsky.social made a fundamental analysis mistake the premise of this article, mixing 2 kinds of Waymo data: - 6 official announcements of having just reached an exact rides/wk milestone, eg 500k/wk - a stmt in February they were doing >400k rides/wk https://techcrunch.com/2026/03/27/waymo-skyrocketing-ridership-in-one-chart/
@prietschka 3 of course just looks like a toy designed to get people hyped up about "agents".
@prietschka I have my doubts that either ARC-AGI-1 or 2 were something you couldn't just brute force train on. Grok for example seems to be deliberately trained on hitting benchmarks that can easily be gamed by just training on the testing data.
@prietschka Quantum computing is in that "shit that ain't gonna happen" category. Not that it's impossible, or can't eventually be useful in some kind of advanced physics research, it's just that there's no commercial use cases, not on the order of selling shit to consumers like you can sell language model tokens for shitty image outputs or shitty copy writing.
https://scottlocklin.wordpress.com/2019/01/15/quantum-computing-as-a-field-is-obvious-bullshit/
RE: https://tldr.nettime.org/@tante/116278661360029668
This is, I think, the story of the last few years. Software is tapped out. There is more software than can be sold. We are in a glut, and we've made it worse by stuffing word embeddings with code that can then be teased out with text prompts. All new software products are solutions in search of a problem, or actively polluting the Web with synthetic data extrusions of dubious quality at best, which threaten to saw off any usefulness of the Web as a technology by turning it into a playground for chatbots.
@prietschka "Higher levels of abstraction" is some meaningless pablum that only programmers could come up with. It's like some leftover 90s-speak when we transitioned from procedural to object-oriented programming.
@prietschka I simply don't understand the point of any of this for consumers. That third and fourth bullet point is insane.