Artificial intelligence across Apple and the industry.
GitHub’s October 22–26 rollout turns eligible unconfigured feature policies into saved settings. Administrators should check individual policies, not rely on a later default change to undo them.
Shared-input caching and 400 questions make Decisions more useful for repeat evaluations. The important limits are typed answers, independent checks and a separate cache policy.
Veo preview endpoints retire October 22. Gemini API developers moving to Omni need to review requests, video output handling, stored state and costs.
Google’s older Deep Research API agent is scheduled to retire October 23. Check the exact identifier, choose a replacement, and test asynchronous completion.
Longer prompt caching can help AI apps reuse context hours apart, but the higher write price makes request timing and actual cache hits decisive.
Gemini can explain the active tab and create static charts on iPhone and iPad. Editing, workbook-wide analysis and plan eligibility need a closer look.
Mac teams can review slides together in Space, then open a deck in Keynote. Check rollout access and decide which copy to keep editing.
Describe and save an app’s voice without recording. GPT-Live-only support, listening costs and retention rules shape deployment.
Meta’s open-weight release highlights memory capacity, distributed inference and the limits of benchmark comparisons.
Mac clients using llama-server may need a new local address after upgrading. Explicit port settings still work, and Docker keeps 8080 inside the container.
