BLEU, or Bilingual Evaluation Understudy, is an automatic metric originally developed for machine translation. It compares short sequences of words in generated text with one or more reference translations, combines precision across several n-gram lengths, and applies a penalty when output is too short. BLEU is inexpensive and useful for comparing systems on the same dataset, but it does not directly measure factuality, fluency, meaning, or human preference. Valid alternative wording can receive a low score, while awkward text can match many reference phrases. Results depend on tokenization and implementation, so reporting should include the exact evaluation settings and, when possible, human judgment.
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.