Loading activity…
Cia lou sen san
AI
The commentary concerns expectations for a possible offering, not a confirmed drop in Anthropic’s valuation.
AI
CoRL 2026 accepted 687 papers from a record 2,094 submissions, a program chair announced.
AI
Figure's robot video drew criticism on X from one viewer, who called its presentation disturbing and potentially dystopian.
AI
The demonstration uses real hardware and is not sped up, according to a post citing Jim Fan.
XAI
Harness reportedly reached the milestone less than three months after launch, counting only official API users.
AI
OpenAI's specialized legal variant (gpt-6-astra-law) reportedly appeared on its API pricing page under a new “Legal model configurations” section.
AI
Google DeepMind's team says it built the 12-degree-of-freedom robot from one plain-English prompt without manual CAD.
XAI
ChatGPT and Codex will share a token-based allowance on OpenAI's Plus, Pro and Business plans; instant chat won't count.
AI
Astra's claimed robotics edge led one observer to speculate about transfer at scale; a reply on X questioned the surprise.
AI
The comparison's author calls fine-grained MoE the best architecture change since GPT-3, but cautions that larger-model estimates are extrapolations.
AI
Qwen reportedly invented Alibaba file-proxy URLs in a separate test, while Cohere says its sandbox blocked an upload attempt.
AI
One invokes a “personal AI agent billboard”; another flips a supposed identity from Claude to Kimi K3.
AI
Neuron Lens's creators report thousands of Gemma 4 12B neurons firing for the same concept across text, images and sound.
XAI
In the preprint's experiments, looping middle layers gave better returns under training-compute limits, while more repetitions could undermine gains.
AI
DeepSeek's open-source Super-Agent prototype can monitor developer group chats and suggest coding solutions without being asked, its team says.
AI
The GLM-5.3 comparison prompted a joke about having different personalities around different friend groups.
AI
A separate post called the option “back,” without explaining why it used that word.
XAI
Random biases to output logits reportedly matched or beat full-weight perturbations on most reasoning tasks tested.
XAI
A tailored index_add_ input reportedly identified A40 chips; the project’s 99.1% result came from a separate custom kernel.
AI
The study compared 64 people with 250 AI-agent trajectories, finding differences in expressed uncertainty and explanations of game mechanics.
