Initial trials indicate CAR T cell therapy, initially for cancer, may reset immune systems in autoimmune conditions, offering hope for new treatments.
Browsing Category
AI & Work
309 posts
OpenClaw creator burned through $1.3 million in OpenAI API tokens in a single month — bill covered 603 billion tokens across 7.6 million requests and 100 coding agents
OpenClaw creator Peter Steinberger’s team consumed over $1.3 million in OpenAI tokens in 30 days, highlighting AI API cost extremes for development projects.
Five times AI hallucinations embarrassed governments
Five instances where AI-generated errors led to government embarrassment, highlighting risks of unchecked AI use in official materials.
The Inference Shift
Cerebras Systems plans to raise its IPO price amid rising AI compute demand, signaling a shift toward heterogeneous AI hardware beyond GPUs.
vLLM V0 to V1: Correctness Before Corrections in RL
Hugging Face reports that vLLM V1 achieved backend parity with V0 after fixing logprob semantics, runtime defaults, weight updates, and fp32 lm_head, prior to RL objective changes.
Unlocking asynchronicity in continuous batching
Explores how asynchronous batching improves GPU utilization by decoupling CPU and GPU tasks, reducing idle time in continuous inference workflows.
SpaceX and Anthropic, xAI’s Two Companies, Elon Musk and SpaceXAI’s Future
Exploring Musk’s involvement with SpaceX, Anthropic, and xAI, and what it reveals about his future plans in AI and space tech.
EMO: Pretraining mixture of experts for emergent modularity
AI researchers introduce EMO, a mixture-of-experts model that naturally develops modular structure during pretraining, enabling selective expert use with minimal performance loss.
The best argument I’ve heard for why AI won’t take your job
A leading SaaS CEO argues AI will augment, not replace, human workers, emphasizing the durability of the last mile of human expertise amid fears of job loss.
Granite Embedding Multilingual R2: Open Apache 2.0 Multilingual Embeddings with 32K Context — Best Sub-100M Retrieval Quality
IBM releases two new open-source multilingual embedding models built on ModernBERT, supporting 200+ languages and code retrieval, under Apache 2.0 license.