AI

Token Superposition Training (TST) by Nous Research speeds up LLM training by 2.5x in 2026. No architecture change needed.

Forum AI, founded by ex-Meta news chief Campbell Brown, evaluates models on high-stakes topics to improve truthfulness.

skfolio library enables building, testing, and comparing modern portfolio investment strategies with Python.

Google and SpaceX are in talks to launch orbital data centers, aiming to make AI compute cheaper in space.

Anthropic share warning: Company says any share sale by unauthorized platforms like Hiive and Forge Global is void.

OpenAI confirms GPT-5 delayed indefinitely, citing diminishing returns from traditional scaling laws.

Exclusive: Perplexity AI's secret ad partnership with major retailers quietly tracks user queries, raising fresh privacy alarms.

OpenAI's o3 model changes AI reasoning, but at what cost? New benchmarks reveal surprising trade-offs in safety and speed.

The GPT-4.1 lawsuit highlights growing legal risks for AI companies using copyrighted data without permission.

Anthropic's Claude 4.5 Sonnet model launches with safety features, but early tests show performance dip. Is it worth it?

OpenAI's Operator AI agent raises serious data privacy concerns as it automates web tasks—potential for mass surveillance

A new jailbreak technique bypasses Claude 4's safety guardrails, exposing critical flaws in Anthropic's alignment strategy.