Good morning. It’s Wednesday, June 25th.

On this day in tech history: In 2013, Instagram launched 15-second video sharing, laying the groundwork for AI-powered content personalization. Backed by cloud-based transcoding, it handled a range of video formats at scale. This move set up the infrastructure for AI-driven video analysis, unlocking smarter ad targeting and content recommendations.

In today’s email:

• Elevenlabs 11ai Assistant

• OpenAI’s Device Lawsuit

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

11ai Automates Slack, Notion, Salesforce, With Just Your Voice

AI productivity assistants are evolving fast, and ElevenLabs is making a serious play for the enterprise market. The company has unveiled 11ai, a voice-first AI assistant designed for execution, not chit-chat. Built around the Model Context Protocol (MCP), 11ai plugs directly into platforms like Notion, Salesforce, and Slack to automate real work: updating CRM records, summarizing documents, assigning tasks, and more, entirely through voice or text.

Under the hood, 11ai uses Retrieval-Augmented Generation (RAG) to ensure responses are context-aware, and it seamlessly switches between voice and text. It supports multilingual conversations with automatic language detection and delivers responses in real time via low-latency processing.

Security is enterprise-ready: HIPAA-compliant with strict permission layers. Users can deploy custom or cloned voices from a library of 5,000+ options to match brand tone.

Rounding out the rollout, ElevenLabs has launched mobile apps for Android and iOS. These apps support content syncing, speech generation in 70 languages, and social media exports. Read more.

OpenAI’s “Third Device” Vision Hits Legal Wall After Trademark Clash

OpenAI is sharpening its hardware and software edge. Court filings from a trademark dispute with audio startup iyO reveal that OpenAI and Jony Ive’s design team are developing a pocketable, environment-aware AI device, neither wearable nor in-ear. Still in prototype phase, the device is at least a year from market.

The legal battle forced OpenAI to drop the “io” brand, which clashed with iyO’s trademark. Internal emails show iyO had repeatedly pitched an acquisition to OpenAI. Sam Altman declined, and later called the lawsuit “silly, disappointing, and wrong,” adding it’s “not cool to sue someone for not buying your company.”

Meanwhile, OpenAI rolled out ChatGPT connectors for Drive, Dropbox, SharePoint, and Box. The feature lets users bring files into ChatGPT for faster analysis and task automation, no Deep Research mode needed. Read more.

• Leak reveals Grok might soon edit your spreadsheets

• Anthropic Scores Fair Use Win for AI, But Piracy Claims Head to Trial

• Google Launches Gemini On-Device Model for Offline Robot Control

• An autonomous AI has just claimed the top spot on HackerOne’s U.S. leaderboard

• New AI by Vibemotion Turns Prompts into Motion Videos

• Microsoft Drops ‘Mu,’ a Fast, Local AI for Talking to Windows

• ChatGPT Nearly Matches TikTok, Facebook, Instagram, and 𝕏 Combined in iPhone App Downloads Over 28 Days

• Sources: Murati Says Thinking Machines Lab Will Build KPI-Tuned AI, Consumer Product Next

• Disney is in talks with OpenAI about possible partnerships involving its characters

• Databricks, Perplexity co-founder pledges $100M on new fund for AI researchers

• Curated realities: An AI film festival and the future of human expression

• AI job postings on LinkedIn grew sixfold as AI skill additions to profiles soared twentyfold

• Beyond static AI: MIT's new framework lets models teach themselves

• Hinge CEO Justin McLeod says dating AI chatbots is ‘playing with fire’

• The music industry is building the tech to hunt down AI songs

• China Reveals Mosquito-Sized Spy Drone for Surveillance Missions

• Japan unveils world's most advanced quantum–classical hybrid computing system

• Pythagora tries to transform vibe coding with specialized AI agent teams

• Scale AI exposed sensitive data about clients like Meta and xAI in public Google Docs, BI finds

• RL-Trained AI Writes 10,000+ Words, No Synthetic Data Required

📄 Light of Normals: Unified Feature Representation for Universal Photometric Stereo

📄 OmniGen2: Exploration to Advanced Multimodal Generation ​​

📄 Drag-and-Drop LLMs: Zero-Shot Prompt-to-Weights

📄 PAROAttention: Pattern-Aware ReOrdering for Efficient Sparse and Quantized Attention in Visual Generation Models

📄 Vision-Guided Chunking Is All You Need: Enhancing RAG with Multimodal Document Understanding

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on 𝕏!