# Morning AI TL;DR — June 11, 2026

- **Anthropic shipped Claude Fable 5 / Mythos 5** — Anthropic says the new models can work autonomously longer than prior Claude versions, with stronger software engineering, vision, memory, knowledge-work, and life-sciences capabilities. ([Anthropic](https://www.anthropic.com/news/claude-fable-5-mythos-5))
- **Cohere open-sourced North Mini Code** — Cohere’s first open-source agentic coding model is a 30B MoE aimed at strong dev-agent performance with lower hardware requirements, useful for teams wanting more control than hosted IDE agents. ([Cohere](https://cohere.com/blog/north-mini-code), [Hugging Face](https://huggingface.co/blog/CohereLabs/introducing-north-mini-code))
- **OpenAI highlighted Codex in real engineering workflows** — OpenAI’s Nextdoor case study frames Codex less as autocomplete and more as a long-running engineering agent for shipping product work. ([OpenAI](https://openai.com/index/nextdoor/))
- **Google DeepMind published AI-learning trial results** — DeepMind reported randomized-trial findings on Gemini guided learning in Sierra Leone, a concrete data point for AI tutors moving from demos to measured education outcomes. ([Google DeepMind](https://deepmind.google/blog/measuring-the-impact-of-learning-with-ai-in-sierra-leone-and-beyond/))
- **EU ordered Meta to reopen WhatsApp access for rival AI assistants** — Regulators told Meta to restore free WhatsApp access for competing AI chatbot providers while an antitrust probe continues, a notable platform-access signal for agent distribution. ([Reuters](https://www.reuters.com/world/eu-regulators-order-meta-allow-rival-ai-chatbots-free-access-whatsapp-2026-06-09/), [AP](https://apnews.com/article/meta-whatsapp-european-union-8b4b48acb23acf5686f141a276bb868d))
- **Agentic coding security risk got sharper** — SecurityWeek reported prompt-injection paths in repo comments affecting Claude Code, Gemini CLI, and GitHub Copilot Agents, reinforcing that coding agents need sandboxing and secret isolation before CI/CD autonomy. ([SecurityWeek](https://www.securityweek.com/claude-code-gemini-cli-github-copilot-agents-vulnerable-to-prompt-injection-via-comments/))
- **AI infrastructure expansion keeps going global** — Meta signed its first AI data-center deal in India with Reliance, underscoring that frontier and enterprise AI capacity is increasingly a geography, power, and partnerships game. ([TechCrunch](https://techcrunch.com/2026/06/10/meta-signs-first-ai-data-center-deal-in-india-with-reliance/))

**Worth watching:** The near-term edge is shifting from “which model is best?” to “which teams can safely wire agents into real workflows, data, and platform channels.”