Why a people list
Agentic AI is moving fast - faster than any documentation site can keep up with. The frameworks (AutoGen, Microsoft Agent Framework, plus a long tail of others) ship breaking changes, the patterns evolve every few weeks, and the actual best practices live in a small number of practitioners' threads, GitHub commits, and conference talks.
The shortest way to stay current is to follow the people who are doing the work. This article is a starter list - a handful of engineers and architects whose feeds are worth turning on notifications for if you are building agentic systems on top of AutoGen or the Microsoft Agent Framework (MAF).
The list
Stephen Toub
Partner Software Engineer at Microsoft. Long-standing voice on .NET runtime, async, and performance work - and increasingly on the agent-framework side of Microsoft's AI investments.
Victor Dibia
AutoGen core developer and Microsoft Agent Framework core contributor. Writes prolifically on multi-agent systems, evaluation, and the practical realities of shipping LLM-driven workflows.
Eric Zhu
AutoGen architect and Microsoft Agent Framework principal architect. The person whose design choices most shape what agents on the Microsoft stack actually look like in production.
How to follow well
A few habits that turn a "people to follow" list into actual signal:
- Read the threads, not just the headlines. The substance is usually in the replies and follow-ups - the original post is the surface, the conversation underneath is where the engineering judgment shows.
- Watch the GitHub activity. Each of the people above commits or reviews in public on AutoGen / MAF. Their pull-request reviews are often more informative than their long-form posts.
- Subscribe to the conference talks. Build, MS AI Tour, AutoGen meetups, and the various open-source conferences are where the longer-form thinking gets shared. The same handful of people show up across all of them.
- Let the list change over time. Three people is a starter, not a canon. As the field grows, swap in whoever is doing the work you care about and drop the names whose focus has drifted.
This page will grow with the path. If you have a name worth adding - someone whose work consistently shifts how you think about agents - tell us, and they will land here.