Google Gemini’s macOS app is testing system-wide voice dictation, cursor tracking, and possible device linking.

Meta launches AI-powered Meta Glasses in partnership with EssilorLuxottica, featuring smart audio, hands-free capture, and 26 frame options.

Anthropic introduces Claude Tag, a Slack agent for teams that transforms Claude into a shared workspace assistant, now in beta.

What's new? OCR 4 extracts document content with boxes, block types and region scores in 170 languages; it is available via API and as a self-hosted container;

OpenAI is rolling out Bidirectional Voice Mode for ChatGPT this week, offering more conversational audio and improved context retention.

Anthropic is testing Cowork for iOS, hinting at cloud-based execution for mobile scheduling and broader access.

Google is developing a "Lit Review" matrix for NotebookLM, designed to organize uploaded sources into a grid for structured research.

OpenAI launches Daybreak updates with Codex Security automation, GPT-5.5-Cyber early access, and Patch the Planet for vulnerability fixes.

What's new? Fugu Ultra is a multi-agent orchestration model available through an OpenAI-compatible API

Perplexity is testing a new shared memory system called Brain, offering users categorized topics, detailed context, and a 3D map for browsing stored knowledge.

What's new? anthropic introduced artifacts in claude code to generate live, shareable visual pages; it is in beta for claude team and enterprise using the cli and app;

What's new? Anthropic launched enterprise-managed authorization for MCP connectors with Okta integration; connectors auto-provision for Claude Chat, Claude Code, and Cowork;

OpenAI is testing real-time voice controls in Codex, hinting at deeper integration with ChatGPT and a unified voice experience.

OpenAI is set to introduce GPT-5.6 model next week, including GPT-5.6-Pro variant, and potentially an updated voice mode.

Microsoft's team behind Copilot Cowork is evaluating a broad range of open models, putting pressure on teams behind proprietary MAI models.

Mistral AI is developing new Code and Apps sections for Vibe (Le Chat) web, while a next-gen model with open weights is set for early access in July.

GLM-5.2 by Z.ai delivers a 1M-token context window for coding agents, project-scale software work, debugging, and code-driven video generation.

What's new? Microsoft announced Copilot Cowork, a cloud-hosted agent system that runs multi tool tasks; it has a usage based billing model and Adobe, Miro and Atlassian plugins;

OpenAI is developing a science-focused ChatGPT plan for research institutions, offering tailored access for universities and R&D teams across disciplines.

Google NotebookLM is testing Personal Intelligence and AI note editing options, hinting at upcoming feature releases.

OpenAI is preparing to launch GPT-Bidi-1, a bidirectional audio model for ChatGPT's voice mode, hinting at a major upgrade in real-time voice conversation.

Mistral has updated its Vibe assistant with a cartoon cat mascot inspired by the viral "Le Chaton Fat" meme, reflecting the company's playful branding shift.

OpenAI’s Codex now gains controlled access to Chrome DevTools, letting it profile JavaScript and modify site elements in its in-app browser mode.

Google Gemini is testing new personalization controls, including support for subscriptions from third-party apps and the ability to manage their context.

xAI is preparing to merge Grok's Tasks feature into a new automations system, offering expanded skill and model selection for scheduled routines.

What's new? Telegram launched smartwatch apps for Apple Watch and Android Wear OS with chat functions; telegram added rich text for bots, ai guardian bots and poll privacy options;

What's new? Anthropic suspended Fable 5 and Mythos 5 following a government export control directive; a bypass method was uncovered during safeguards testing;

What's next? Google is testing the integration of Skills Marketplace inside Gemini Enterprise, and more.

What's new? MiniMax M3 is a multimodal model on NVIDIA accelerated compute with text, image, video support and sparse attention for long tasks available via NVIDIA API;

Meta AI is testing Deep Research, Presentation, and Social modes on the web, expanding beyond chat to offer research, deck creation, and other options.