This week’s notes focus heavily on the practicalities of building reliable AI agents, specifically looking at how we manage their memory and governance. I have also included a few notable wins in Azure storage performance and some honest reflections on where our industry is heading next.
AI
AI Governance: Runtime Enforcement, Not Runtime Advice (Karan Verma) - I'm impressed by our journey of using AI. Not long ago, we barely trusted it, and now it's running independently, and we think more about limiting what it can do. Nice second post o to a series.
Computer use: How AI agents can automate almost anything | Red Hat Developer - Learn more about the different options that Red Hat AI offers and how they integrate with everything you may already have.
Native Agent Memory for Microsoft Agent Framework, Powered by Azure Cosmos DB - Azure Cosmos DB Blog (Theo van Kraay) - That seems pretty cool! Attaching a serverless NoSQL database to our agent to get super memory.
Squad 0.11.0: portable teams and a harder release train · Brady Gaster (Brady Gaster) - I have used Squad from the beginning, and it's great to see it evolving. Thank you for this update post! I wish the release notes (of all projects) were that clear sometimes.
I've Spent 25 Years in Tech. The AI Gap That Worries Me Most Isn't the One You're Thinking About (Fabrice Lacroix) - Great reflection on AI for our business. Are we ready?
Cloud
- Faster blob copying in Azure Blob Storage (Mark Heath) - Some may think it's trivial, but when you hit big numbers, the way you copy makes a huge difference.