# Morning AI TL;DR — May 11, 2026

Fresh, high-signal AI news was fairly light over the weekend, so this includes the most relevant updates from the last ~48 hours plus a few strategically important May 5–6 releases.

- **OpenAI made GPT-5.5 Instant the ChatGPT default** — OpenAI says the new default improves factuality, clarity, personalization controls, and everyday technical work, which matters for builders relying on ChatGPT as a baseline assistant. ([OpenAI](https://openai.com/index/gpt-5-5-instant/), [release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes))
- **U.S. pre-release AI model review is getting real** — Google DeepMind, Microsoft, and xAI reportedly agreed to let the Commerce Department’s CAISI evaluate models before public release, a notable shift toward operational safety review rather than pure voluntary self-reporting. ([The Verge](https://www.theverge.com/ai-artificial-intelligence/924017/google-microsoft-xai-government-review), [NIST CAISI](https://www.nist.gov/caisi))
- **Anthropic raised Claude limits and added SpaceX compute** — Anthropic says higher usage limits and a SpaceX Colossus 1 capacity deal should expand near-term Claude availability, especially relevant for Claude Code and agent-heavy workflows. ([Anthropic](https://www.anthropic.com/news/higher-limits-spacex), [The Verge](https://www.theverge.com/ai-artificial-intelligence/925348/anthropics-claude-usage-limits-are-getting-a-boost-after-compute-deals-with-spacex-and-others))
- **Google’s Gemini CLI is pushing terminal-native agents mainstream** — Google’s open-source Gemini CLI brings Gemini into the terminal and shares tech with Gemini Code Assist, continuing the trend from chat copilots to command-line coding agents. ([Google](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-cli-open-source-ai-agent/), [GitHub](https://github.com/google-gemini/gemini-cli))
- **Salesforce keeps moving Agentforce deeper into vertical workflows** — Merck Animal Health selected Agentforce Life Sciences/Agentforce 360 as a “OneCRM,” showing Salesforce’s agent strategy shifting from demos to regulated, industry-specific customer-engagement systems. ([Salesforce Investor Relations](https://investor.salesforce.com/news/news-details/2026/Merck-Animal-Health-Selects-Salesforces-Agentforce-Life-Sciences-for-Customer-Engagement-to-Transform-Animal-Care-and-Enhance-Support-Experiences-for-Employees/default.aspx))
- **Agentforce Vibes adds intent-driven tool selection** — Salesforce’s developer blog describes “Abilities” for deciding which tools an agent should access at runtime, useful for anyone designing safer, more contextual enterprise agents. ([Salesforce Developers](https://developer.salesforce.com/blogs/2026/05/intent-driven-tool-selection-using-abilities-in-agentforce-vibes))
- **OpenCode remains one to watch in open-source coding agents** — recent indexes describe OpenCode as a provider-agnostic alternative to Claude Code/Copilot with desktop beta support, reinforcing demand for portable agent tooling. ([OpenCode](https://opencode.ai/), [GitHub](https://github.com/anomalyco/opencode/))

**Worth watching:** Enterprise agent adoption is likely to hinge less on model benchmarks and more on governed tool access, compute availability, and integrations into systems of record like Salesforce.

**Artifact:** `/home/tjonestj3/.hermes/hermes-agent/outputs/ai-news/2026-05-11-ai-news-tldr.md`