Hume.ai released EVI 3, a new personalized voice AI model

Hume.ai launched EVI 3, a third-gen speech-language model combining trannoscription, reasoning, and speech synthesis with rapid response and custom voice creation. It outperforms peers in expressiveness, empathy, and natural voice delivery.

🗞 #ai

Mistral AI opens Agents API for public use with task planning and tool integration

Mistral AI launched the Agents API via La Plateforme, allowing developers to build autonomous agents that handle tasks, integrate tools, and retain context. It supports coordinated workflows and emphasizes EU data residency.

🗞 #mistral

Upcoming Claude feature lets users explore and share AI-generated Artifacts

Anthropic is launching the Artifacts Gallery in the Claude web app, showcasing user-created outputs like code, diagrams, and mini-apps. Users can browse, customize, and share artifacts, supporting discovery and collaborative AI use.

🗞 #claude

Users can now watch Perplexity AI control pages in Comet Browser

Comet Browser introduces a visual side chat bar enabling users to direct Perplexity AI on active pages, expanding from headless automation to real-time, transparent control for complex tasks across multiple tabs and workflows.

🗞 #perplexity

xAI readies image discovery tool for Grok Web platform

xAI is adding new features to Grok Web, including a decorative idle animation, Google Calendar integration for scheduling support, and an image exploration tool. These updates align with its goal to position Grok as a versatile AI assistant.

🗞 #grok

Microsoft tests new Copilot Live Portraits feature with customizable avatars

Microsoft is testing a Copilot feature called Live Portraits, which may introduce customizable visual avatars for voice-based use. Still in early development, it reflects a broader move toward a personalized, avatar-driven assistant interface.

🗞 #microsoftcopilot

Comet, Agentic AI browser by Perplexity: everything you need to know

Comet is Perplexity’s new Chromium-based AI browser with built-in automation, native tab and history access, contextual commands, and Gmail/Calendar integration, offering hands-off browsing and intelligent task execution from multiple entry points.

🗞 #perplexity

OpenAI preparing custom ChatGPT connectors via Model Context Protocol

OpenAI is expanding ChatGPT's connector support with the upcoming MCP integration, allowing users to add custom services and endpoints. Aimed at business use, this feature enables tailored access to external systems and enterprise tools.

🗞 #chatgpt

New Claude capability in the works to merge Research and MCP integrations

Anthropic is merging Claude’s research mode with MCP support into a single feature for Claude Max users, enabling tool integration, extended reasoning, and automation—positioning it as a productivity-focused solution for enterprise use.

🗞 #claude

OpenAI to release o3 Pro for ChatGPT Pro users today

OpenAI is launching the o3 Pro model for ChatGPT Pro users, expanding advanced reasoning capabilities and integration features for business use. Updates include broader support for third-party apps and custom endpoints via the MCP framework.

🗞 #chatgpt

Mistral Code launches in private beta for JetBrains and VS Code users

Mistral AI has launched Mistral Code, an enterprise-grade coding assistant supporting over 80 languages, with code completion, refactoring, and AI model customization, targeting secure, on-premise deployments and regulatory-conscious sectors.

🗞 #mistral

Character.AI shifts to media-rich platform with Scenes and social feed

Character AI expands beyond text-based chats with multimedia tools such as avatar video creation, animated scenes, dynamic character conversations, and a centralized content feed, supporting deeper storytelling across its web and mobile platforms.

🗞 #characterai

Google’s Kingfall model briefly goes live on AI Studio before lockdown

Google's confidential Kingfall model briefly surfaced on AI Studio, featuring multimodal input, advanced SVG and code generation, and a deeper “thinking” mode. Its leak hints at a potential Gemini 2.5 Pro release too.

🗞 #aistudio

Google to bring Canvas upgrades and Enterprise mode to Gemini

Google is adding Enterprise mode for professional use. New features support video from images and automatic creation of visuals like infographics, timelines, and presentations.

🗞 #gemini

Upcoming Grok’s Tasks feature gets UI upgrade and scheduling tweaks

xAI's Grok introduces a sidebar for easier access to its developing Tasks feature, adds flexible notification settings, and refines scheduling, aligning with industry trends toward automated, scheduled AI actions for power users and business applications.

🗞 #grok

New Genspark feature helps users manage meetings, emails, and notes

Genspark AI Secretary is a new tool for professionals and small teams, offering task management, scheduling, note-taking, and reminders through voice or text. It integrates with productivity tools and is accessible via the Genspark web platform.

Sparkify uses Gemini and Veo 3 to power Google’s latest AI video experiment

Google's Sparkify is an invite-only AI video creation tool using Gemini and Veo to generate short, thematic videos in various styles. It targets education and entertainment, with limited interactivity and no confirmed release timeline.

🗞 #aitestkitchen

Comet AI browser agent now captures screenshots during tasks

Perplexity's Comet browser now includes screenshot capture for improved AI agent transparency and task validation. Outlook support and a Windows release are in progress, with mobile deployment set for fall to support cross-platform automation.

🗞 #perplexity

Google expands Project Mariner access to more Ultra subscribers

Google's Project Mariner, a browser assistant for Ultra subscribers, grants tab-level access in Chrome via extension. Still experimental, it faces limitations due to frequent permission prompts and reliance on manual input.

🗞 #aitestkitchen

Mistral AI debuts Magistral models focused on advanced reasoning

Mistral AI launched the Magistral model family, featuring the proprietary Magistral Medium and open-source Magistral Small, targeting advanced reasoning tasks with a focus on performance, cost efficiency, and flexible deployment options.

🗞 #mistral

Google Cloud debuts G4 VMs optimized for large-scale AI inference

Google Cloud's new G4 virtual machines, powered by eight NVIDIA RTX PRO 6000 Blackwell GPUs and AMD Turin CPUs, offer quadrupled computing and memory capacity over G2 VMs, targeting AI, simulation, and graphics-heavy enterprise workloads.

🗞 #ai