Latency measures the delay between an AI request and the resulting response. Interactive generative systems often track time to first token, which reflects how quickly output begins, and time to last token, which captures total completion time. Latency comes from model computation, queueing, network transfer, retrieval, tool calls, safety checks, and application logic. It varies with prompt length, output length, model size, hardware, batching, cache use, and current load. Optimizing only average latency can hide poor experiences, so teams also monitor percentile values and separate each stage of the request. The right target depends on whether the product is conversational, analytical, offline, or safety-critical.
AI tarot apps and chatbots are competing with human psychic services. Some providers now advertise readings without AI as a selling point.
University of Kentucky administrator David Durman secured an exemption from personal AI use, citing humanist beliefs about independent thought and responsibility.
Claude Code users report a rule against recreating existing games. Original projects remain allowed, but a licensed developer says their work was refused.
Anthropic’s Linux exploit reached root in 53 minutes on an exposed Multipath TCP server. The 9.8-rated flaw is fixed upstream, but vendor rollout varies.
Playgama’s AI Game Index compares coding tools and models through player engagement across 5,400-plus browser games, with rankings updated daily.
Mistral opens Le Chonk’s weights, claims coding parity with GLM-5.3, and lists API rates of $0.70 for input and $2.10 for output per million tokens.
OpenAI doubles limits for Pro and Business Premium, cancels a planned Pro 200 Work and Codex cut, and keeps extended dots usage separate.
Dario and Daniela Amodei discuss leaving OpenAI, AI deception, military power and meaning after AGI in Walter Isaacson’s newly released interview.
Michael Selig imagines a crypto-funded fridge and asks who is responsible when AI agents trade, spend or act beyond their owner’s instructions.
NVIDIA’s Nemotron 3.5 Transcribe handles hour-long recordings in 40 languages, including Russian, with speaker labels, translation and English audio Q&A.