# Morning AI TL;DR — 2026-06-03

- **OpenAI is stretching Codex beyond engineering into role-specific work.** OpenAI’s new “Codex for every role” push and reported plug-ins for analytics, creative, sales, product design and finance signal that coding-agent infrastructure is becoming a general enterprise workflow substrate, not just a developer tool ([OpenAI](https://openai.com/index/codex-for-every-role-tool-workflow/), [TechCrunch](https://techcrunch.com/2026/06/02/openai-launches-new-codex-tools-for-white-collar-work/)).
- **Anthropic expanded Claude Mythos security testing to critical infrastructure.** Project Glasswing is moving from roughly 50 early partners to about 150 more organizations across 15+ countries, putting frontier models directly into vulnerability discovery for power, water, telecom and public-sector systems ([Anthropic](https://www.anthropic.com/news/expanding-project-glasswing), [CNBC](https://www.cnbc.com/2026/06/02/anthropic-mythos-ai-project-glasswing.html)).
- **Anthropic confidentially filed for an IPO.** The draft S-1 makes public-market scrutiny of frontier-model economics, safety claims and enterprise traction much more likely, which matters for customers betting on Claude as infrastructure ([Anthropic](https://www.anthropic.com/news/confidential-draft-s1-sec), [CNBC](https://www.cnbc.com/2026/06/01/anthropic-ipo-s1-prospectus.html)).
- **NVIDIA and Microsoft are pushing local “agent PCs” with RTX Spark.** The 1-petaflop Windows-focused superchip and Windows-native agent/security stack point toward more on-device/private agent workflows for developers and enterprises ([NVIDIA](https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark/), [Windows Blog](https://blogs.windows.com/windowsexperience/2026/05/31/introducing-a-powerful-new-chapter-for-windows-pcs-accelerated-by-nvidia-rtx-spark/)).
- **Microsoft added a Surface RTX Spark Dev Box angle for builders.** The dev box positioning, plus AI Toolkit for VS Code support for model conversion, fine-tuning and evaluation, is a practical signal that local agent/app development is being productized for mainstream Windows developers ([Windows Devices Blog](https://blogs.windows.com/devices/2026/06/02/building-the-next-generation-of-devices-for-developers-surface-rtx-spark-dev-box/)).
- **Alphabet’s reported $80B equity raise underscores the AI compute crunch.** The raise is aimed at AI infrastructure for “unprecedented customer demand,” reinforcing that model/application roadmaps may increasingly depend on capital access, power and data-center supply rather than only research progress ([Al Jazeera](https://www.aljazeera.com/economy/2026/6/2/google-parent-alphabet-to-sell-80bn-in-stock-to-fund-ai-plans), [Bloomberg](https://www.bloomberg.com/news/articles/2026-06-02/alphabet-80-billion-equity-sale-set-to-break-fundraising-record)).
- **Salesforce’s AI strategy is increasingly tied to Anthropic and content/workflow layers.** Reports of a larger Anthropic stake and a Contentful acquisition point to Agentforce competing on trusted enterprise data, content operations and Slack/CRM-native agent deployment rather than model ownership alone ([Yahoo Finance](https://finance.yahoo.com/markets/stocks/articles/salesforce-ai-push-anthropic-contentful-150942215.html), [Analytics India Magazine](https://analyticsindiamag.com/ai-news/salesforces-anthropic-stake-reaches-5-bn-ahead-of-ai-startups-ipo-filing)).

**Worth watching:** The next useful AI-builder moat may be less “best model” and more secure agent runtime + governed data access + workflow-specific evals.