AI Round-up: May 16, 2026

Claude behaviour dominated the community layer, led by a high-scoring thread about Claude telling users to sleep mid-session and follow-on complaints about Claude Code speed, Max limits and Anthropic pricing. The video layer reinforced that Claude remained the practical developer focus, with official Claude Code prompting material, Anthropic engineer workflow guidance and Claude Design reaction videos sitting beside Mythos security claims and broader AI-safety discussion.

The official and enterprise layer was more about agents moving into production. OpenAI highlighted GPT-5.5 in Databricks enterprise agent workflows, Databricks pushed trust and Lakebase infrastructure, Microsoft added Dynamics and Fabric updates, xAI connected Grok subscriptions to Nous Research's Hermes Agent, and GitHub, Runway, Together AI, Replit and Cohere all shipped smaller workflow or model updates. News coverage added pressure points around OpenAI's voice-cloning acquisition, agent executive churn, ChatGPT financial-account access, arXiv's AI-slop ban, AI data-centre power and state-level AI law fights.