Meta says Muse Ember matches Claude Haiku 5.5 at about a third of the cost per task
Same per-token price as Claude Haiku 5.5 and GPT-6 Luna on short prompts: Meta is betting on the cost of each task.
Same per-token price as Claude Haiku 5.5 and GPT-6 Luna on short prompts: Meta is betting on the cost of each task.
Job cuts blamed on AI fell from 38,579 in May to 1,265 in October. Payroll studies point to slower hiring of young workers.
Google agent teams cut quantum program costs by up to 290x, proved each result in Lean and swapped ideas at scheduled conferences.
Amazon’s HOPE removes half the experts from an open Qwen model. Its agentic coding score holds, but general benchmarks slip, the paper’s tables show.
Claude Fable 5.2 costs 20% less on the API, and Pro subscribers can use it at no extra cost until November 16.
Trained on 300 hours of simulated data, a Unitree G1 passed 91% of 360 single-skill trials in real homes. Chained tasks averaged 70%.
Google amended its 2023 promise: humans may now review excerpts of emails connected to Gemini. The EU is excluded for now.
Fuji TV says AI cut work hours by up to 70 to 80% on some steps. It plans AI for every drama around October 2027.
Open WebUI now links its users’ computers to AI through Tether. Businesses need an Enterprise License, and there is no public price.
Anthropic explains how Claude agents keep notes between sessions, and when the “dreaming” pass that corrects them actually improves results.