# Morning AI TL;DR — 2026-06-19

- **OpenAI + Boston Children’s used o3 Deep Research to surface 18 rare-disease diagnoses** from 376 previously unsolved pediatric cases in an NEJM AI study — practical evidence that reasoning models can augment specialist workflows when paired with clinicians. [OpenAI](https://openai.com/index/diagnose-rare-childhood-diseases/) / [NBC News](https://www.nbcnews.com/tech/innovation/ai-boston-childrens-hospital-diagnose-rare-diseases-kids-openai-rcna350387)
- **Z.ai’s GLM-5.2 is getting attention as a strong open-weights coding/agent model** with a reported 1M-token context window and MIT-licensed weights — worth testing for long-repo coding agents and private deployment paths. [Simon Willison](https://simonwillison.net/2026/Jun/17/glm-52/) / [DataCamp](https://www.datacamp.com/blog/glm-5-2)
- **Amazon is reportedly exploring direct sales of Trainium AI chips** to third-party data centers, not just offering them through AWS — a meaningful move toward more non-NVIDIA accelerator supply and potentially lower inference/training costs. [TechCrunch](https://techcrunch.com/2026/06/18/amazon-hopes-to-challenge-nvidia-more-directly-by-selling-its-ai-chips/) / [Bloomberg](https://www.bloomberg.com/news/articles/2026-06-18/amazon-is-in-talks-to-sell-nvidia-rival-chips-to-other-companies)
- **Anthropic added enterprise-managed MCP connector access, starting with Okta,** bringing centralized authorization across Claude chat, Claude Code, and Cowork — a concrete step toward governable enterprise agent rollouts. [Releasebot](https://releasebot.io/updates/anthropic/claude) / [Claude blog](https://claude.com/blog)
- **Salesforce’s $3.6B Fin deal remains the big CRM-agent ecosystem story this week,** aimed at strengthening Agentforce customer-service and agent customization capabilities — relevant for builders betting on CRM-native agent workflows. [CRN](https://www.crn.com/news/ai/2026/salesforce-targets-3-6b-fin-deal-to-enhance-agentforce-ai) / [TechRepublic](https://www.techrepublic.com/article/news-salesforce-acquires-fin-agentforce-ai-agents/)
- **OpenAI’s ChatGPT release notes now flag upcoming model retirements,** including GPT-4.5 in ChatGPT on June 27 and o3 on August 26 — teams should audit pinned workflows and evals before defaults shift. [OpenAI Help Center](https://help.openai.com/en/articles/6825453-chatgpt-release-notes)

**Worth watching:** The highest-leverage near-term bet is testing GLM-5.2 or similar long-context open models against real coding-agent tasks while enterprise MCP/auth patterns mature for governed deployment.