WWDC set — Siri overhaul
Apple confirmed WWDC for June 8–12 and says it'll stop chasing a single chatbot in favor of embedding AI across iPhone, iPad, Mac and other products (June 8–12). (techlusive.in) (timesnownews.com). The centerpiece: a major Siri overhaul that may include an AI App Store, third‑party chatbot integrations and a standalone Siri app — and Ollama is already using Apple's MLX framework to speed local AI on Macs, which DevOps teams say smooths local testing and workflows. (analyticsinsight.net) (techstory.in) (x.com)
Bloomberg reports Apple will let Siri route queries to third‑party AI assistants such as Google’s Gemini and Anthropic’s Claude via a new “Extensions” system, according to Mark Gurman (March 26, 2026). (bloomberg.com) Gurman’s reporting says Apple expects the Extensions approach to let third‑party AI subscriptions flow through the App Store as a revenue channel, and that the system is distinct from Apple’s separate Gemini partnership. (bloomberg.com) Leaks and reporting describe a dedicated, chat‑style Siri app that can pin past conversations, accept uploaded documents and photos for analysis, switch between text and voice modes, and surface an “Ask Siri” system toggle in the UI. (gadgets360.com) Ollama’s GitHub changelog shows the project merged an MLX runner and related fixes into its v0.19.0 release candidates, adding support for MLX-specific snapshots and new import formats such as mxfp4/nvfp4. (github.com) Coverage from The New Stack and Apple’s ML research notes explain MLX’s shared‑memory model on Apple Silicon reduces CPU–GPU transfer overhead and improves inference responsiveness and throughput for local LLMs. (thenewstack.io) Third‑party previews and community tests of Ollama’s MLX build report concrete speed gains — a preview cited roughly 1.6× faster prefill and nearly 2× faster decode on Apple Silicon, with the biggest improvements noted on M5‑class hardware and higher unified‑memory configurations. (letsdatascience.com)
