# Morning AI TL;DR — June 6, 2026

- **OpenAI upgraded ChatGPT memory with “Dreaming.”** The new system is designed to keep preferences and context fresher across conversations, a practical step toward persistent personal/work assistants. [OpenAI](https://openai.com/index/chatgpt-memory-dreaming/)
- **Microsoft launched MAI-Code-1-Flash for Copilot/VS Code workflows.** Microsoft says the in-house coding model is optimized for fast everyday agentic development and built on clean/appropriately licensed data, signaling more model diversification inside developer tools. [Microsoft AI](https://microsoft.ai/news/introducingmai-code-1-flash/), [CNBC](https://www.cnbc.com/2026/06/02/microsoft-unveils-new-ai-models-lessen-reliance-on-openai-lower-costs.html)
- **Meta’s developer model API reportedly slipped again.** Reuters says Meta has repeatedly delayed the Muse Spark API, underscoring that open/developer model roadmaps remain execution-risky even for hyperscalers. [Reuters](https://www.reuters.com/technology/meta-repeatedly-pushes-back-new-ai-model-release-developers-wsj-says-2026-06-04/)
- **NVIDIA pushed agentic AI deeper onto Jetson edge devices.** JetPack 7.2 brings Jetson Linux 39.2, Ubuntu 24.04 base, CUDA 13, Yocto support, and NemoClaw support—useful for robotics, factories, and local/secure agents. [NVIDIA Developer Forums](https://forums.developer.nvidia.com/t/jetpack-7-2-jetson-software-goes-agentic-with-jetson-linux-39-2/372056), [NVIDIA Newsroom](https://nvidianews.nvidia.com/news/latest)
- **OpenAI expanded controlled life-sciences access via Rosalind Biodefense/GPT-Rosalind.** Access is framed around vetted organizations and public-health/biodefense use cases, showing frontier models moving into high-stakes vertical workflows with tighter governance. [OpenAI research listing](https://openai.com/research/index/release/), [Moneycontrol](https://www.moneycontrol.com/news/trends/openai-launches-rosalind-biodefense-to-boost-pandemic-readiness-13937462.html)
- **The White House issued a new AI innovation/security action.** The order emphasizes private-sector collaboration, securing AI-enabled systems, and protecting IP—policy that could shape procurement and compliance requirements for enterprise AI builders. [White House](https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/)
- **Salesforce’s AI roadmap remains centered on Agentforce + enterprise workflows.** Recent Summer ’26 and Agentforce/partner updates point toward governed agents embedded in CRM, Slack, data, and back-office operations rather than standalone chatbots. [Salesforce Summer ’26](https://www.salesforce.com/news/stories/summer-2026-product-release-announcement/), [Salesforce + Anthropic](https://www.salesforce.com/in/news/stories/salesforce-anthropic-trusted-context-ai-actions-on-claude/)

**Worth watching:** The next competitive wedge is less “which chatbot is smartest” and more which platforms can safely persist memory, execute actions, and run agents across enterprise data without blowing up governance.