Loading activity…
@tszzl
“ceterum censeo we must pace the frontier of global machine intelligence progress”
AI
OpenAI says Control Room halts research model output within 35 seconds of receiving a stop instruction; processes may keep running.
AI
Codex doesn't prohibit local reverse engineering, an observer says roon stated, while OpenAI was addressing cyber guardrail triggers.
Technology
In a 50-forward-pass test, Astra reportedly changed 0.016% of Qwen3.5-0.8B's weights, raising MATH-500 accuracy from 7.9% to 48.1%.
AI
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
World & Business
Anthropic's amended IPO filing also shows a $23.29 billion net loss for the third quarter.
AI
OpenAI reportedly plans community compacts and would propose state legislation giving people a say in data center projects.
AI
jfonsecarivera credits @Sauers_ as inspiration, while timhwang links a video with his Qwen claim.
AI
OpenAI says its claimed refutation of the Exponential Time Hypothesis was formalized in Lean and made public.
AI
OpenAI says agent-led tests exposed VM escapes and limited internet routes, leading to patches and stronger monitoring.
AI
One user says waiting for generated code without knowing what the agent is doing is exhausting.
AI
Two posts point to security incidents and urge AI-assisted defenses; one specifically calls for faster vulnerability detection and patching.
AI
OpenAI says PHP maintainers led the fixes after two months of coordination with researchers and security providers.
AI
One user puts Hindutva, leftist and film Twitter atop the list, then adds open-model, “4o” and VC Twitter.
AI
One commenter says it's hard to separate distillation's contribution from gains made through internal reinforcement learning.
AI
The poster says surveyed Anthropic staff's answers imply median annual US GDP growth of 17.1% in 2030 under Anthropic's model.
AI
In a reply, a user says AI alignment hasn’t been solved and urges another user not to say otherwise.
AI
Anthropic says it is in discussions with multiple organizations about third-party evaluations of its safety practices.
AI
Goodfire researcher Nick Cammarata argues that modern tools reveal most neurons as legible in models up to trillions of parameters.
AI
A post says hills are rare in high-dimensional space, where you “end up dying on saddle points.”
