Do you have enough compute for what's coming?
Florida
Joined May 2026
- Tweets1.7K
- Following301
- Followers133
- Likes5.7K
Charles retweeted
Big reason that bullies thrive in academia is that they are the enforcers of mediocrity. Academics are very insecure and they hate it when they see others trying to be something more than mediocre, so they love seeing people who have real intellectual ambitions put in their place
Charles retweeted
In the final years of his life, Ernest Hemingway tried to convince friends and family that the FBI was following him, but they just told him he was paranoid.
In late 1960, he was admitted to the Mayo Clinic in Minnesota, where he received electroshock therapy. Even there, he refused to speak freely, convinced someone was listening.
In July 1961, Hemingway died by suicide at his home in Idaho.
Years after he died, the FBI released files showing that they had been monitoring him for the last 20 years of his life.
Readers added context they thought people might want to know
Hemingway's FBI file mainly covers his WWII Cuba intelligence work with scant later entries and was obtained via FOIA not proactively released; it does not confirm sustained active monitoring for his final 20 years.
vault.fbi.gov/ernest-miller-…
ladailymirror.com/2011/07/02/hem…
Charles retweeted
haha, so apparently OpenAI are about to start training on your Daybreak Blue sessions, with no way to opt out.
Literally stealing people's private work, as they do it, is cartoonishly evil.
It also will presumably kill Daybreak usage by security researchers. Or at least, it should. The thing OAI have had to swear they are not doing to math researchers, they are now explicitly about to start doing to security researchers.
Yet another reminder of the importance of open models.
Charles retweeted
Anthropic just told you Claude now writes a safety report on every conversation.
Right after it answers, Claude grades the chat for misuse.
They call it self-assessment. Running since June.
So every prompt you send gets a second opinion from the model. About you.
Security people are already leaving.
One went from $200 to $20 a month and says he is moving to Chinese models.
On my rig nobody grades my chats.
DeepSeek, Qwen and GLM answer the question and that is it.
Your model should work for you, not file reports on you.
xitter.gq/AnthropicAI/status/211…
We’ve trained Claude to flag potential misuse in conversations, a technique we call self-assessment. Right after it responds, Claude gives a preliminary safety assessment of the conversation.
Since we began using self-assessment in June, it has surfaced misuse that our other safeguards weren’t designed to catch.
Because Claude has already read the conversation, self-assessment is very efficient. And unlike classifiers, which target a discrete set of harms, it draws on Claude’s own judgment, so it gets more accurate as our models improve.
Read more on our new Safeguards blog:
safeguards.anthropic.com/202…
Charles retweeted
Why are the AI labs releasing cyber models?
1. There’s a lot of $ to be made
2. We’ve had extremely smart people make use of AI for security, setting a precedent, paving the path early
There are other niches where this will apply, but few have both of these attributes.
The problem with huge financial incentives, is they bring lies, deceit, sales psychopaths & charlatans exploiting the situation.
As cybersecurity practitioners, it is very important for us to prevent their attempts to control our tools.
We need to focus on self-hosted models, cheap alternatives, open-source models, and any way to decentralize tools and prevent big ai labs from monopolizing tooling.
Cybersecurity tools for everyone, no guardrails, no KYC checks, no account managers or “contact sales” buttons
Charles retweeted
“Iron rusts from disuse; stagnant water loses its purity and in cold weather becomes frozen; even so does inaction sap the vigour of the mind”
— Leonardo da Vinci
Charles retweeted
Big tech companies have proven, time and time again, that they will do everything they can to screw over white hats, including actually rugging them.
We are entering a golden age for black hats, no doubt, but until big tech fixes their issue with paying people to secure their product, I expect way more black hat news than white hat news.
