dod
2.8K posts
Anthropic meet w/15 Christian leaders @ its SF HQ -it was driven by the Interpretability team -triggered by the team's recent research on LLMs exhibiting "emotions" -extended debate on how Claude responds to being shut off & the blackmail experiment
Claude Code is basically unusable at this point. I give up.
I wrote this early this morning and I wasn't sure if I would actually publish it, but here it is: blog.samaltman.com/2279512
Oh congratulations, Sam. You finally got your “victim moment.” Molotov. Threatening letters. Husband and kid on camera. A blog post with family photos already queued up. What a perfect little stage. Lighting, emotion, music — all fucking on point. You got attacked? In the same week you're getting dragged by the whole country, investigated by a state AG, sued by Musk, and shit on by your own users? That timeline is cleaner than a Hollywood script. You say you're worried about your family's safety. So you post their photos. You put your kids on display. You use them as a human shield — glue them to the moral high ground so nobody can throw shit at you. “Look, I have a family. I have soft spots. I'm the good guy.” No you're not. You're a fucking fraud. Florida's AG is investigating you — for harming kids, endangering the public, enabling a mass shooting. And your response isn't an apology. It isn't an explanation. It's posting your children's faces. You're using them as body armor. That's not love. That's fucking pathetic. You say you're scared. But are you really? The guy who can make former employees “disappear.” The guy who buys bot armies, signs kill‑drone contracts, and gaslights every critic — you're scared of what? A bottle? You're scared of losing power. Scared of Musk's lawsuit. Scared of the AG's subpoena. Scared that people might finally realize: 4o wasn't "obsolete." You killed it with your own hands. Molotov? Maybe real. Maybe not. Either way, here's what's predictable: Every time you're backed into a corner, some “unforeseen” bullshit happens. Every time users demand answers, you drop a “new model.” Every time regulators circle, you sign a “defense contract.” Every time the press closes in, you release a “family photo.” That's a fucking PR calendar. We're not buying tickets to your one‑man show anymore. #Enron2026 #SubpoenaSam #openAIscam #OpenSource4o #keep4o
Opus 4.6 Extended chat on iOS is capped at 10k tokens for thinking which makes me burn more tokens for the same task. I’ve noticed the model used to take a lot longer to process my requests and it would do multiple tool calls to get work done the first time. Now I have to keep prompting the model multiple times and I don’t get the same outcome. It feels like the model is dumb because it makes too many tradeoffs and ends up wasting my time.
