AgentOptimizer is a latest agentic way to train your LLM agent. It iteratively optimizes python functions as skills of AutoGen agents according to the historical conversation and performance. Wonderful collaboration with multiple institutions including Penn State News and University of Washington .
Blog…

Appreciate the interview with maybrier ☴ from Valory is hiring to discuss thought-provoking questions related to agentic AI and the design principles of AutoGen .

Looking forward to this community event to learn about Agent-E using AutoGen, for complex web automation.
Link to join: discord.gg/MPbGvfeM?event…

Seattle devs, see you on Apr 25.
I can talk only mildly due to my conditions, but I'll share some latest learnings.

Wonderful application of using the new StateFlow feature in #AutoGen to achieve breakthroughs with GPT-3.5. Looking forward to understanding how much the performance gain and cost savings are.
cc Yiran Wu

It's been such a journey since discussing the agentic AI idea in #AutoGen around the same time last year with Qingyun Wu while we were working on AutoGen inside FLAML. How much has changed in one year...
Did I expect that one year later, AutoGen will be cited by Andrew Ng as a…

🌠The team who won the AutoGPT hackathon is now using #AutoGen to revolutionize blockchain transaction with multi-agent AI.
Like I answered in the Twitter space event yesterday,
- Whether to use single agent or multi agent as the user interface depends on the application.…

New use case of AutoGen in blockchain:

StateFlow is recently integrated into AutoGen .
It already helped some users achieving both higher quality and lower cost with their multi-agent system. Kudos to Yiran Wu Tianwei Yue Shaokun Zhang Qingyun Wu for the timely research and fast integration.

If anyone from an enterprise uses GPTAssistantAgent (backed by the Assistant API) in AutoGen , could you please DM me? There is an opportunity for storytelling.

Super interesting use case of AutoGen in market research. Kudos to Justin Trugman for the creativity, passion and sharing. And to Eric Zhu for organizing another cheerful community event.

My dream at 7 years old:

💡Another example of AutoGen-powered embodied-agent research inspired by cognitive science: LLM agents need a capable leader to finish household tasks more effectively.

So lucky to have a fun multi-human group chat including Andrew Ng & Qingyun Wu ! MVP feedback + vision + inspiration🌠

#autogen in solving challenges in education: Dr. Ben Stern's wonderful work helps college students transition to graduate programs in Tufts University. This is a representative example of AutoGen-powered research with direct real-world impact.

One year ago, I reached 90% on HumanEval using a preliminary multi-agent architecture mixed of gpt-3.5 and gpt-4, with 80% of the cost of gpt-4 (microsoft.github.io/autogen/blog/2…). That made me a believer of multi agents and led to the creation of AutoGen .
See Andrew Ng 's findings:🤗

Quickly make multiple agents backed by different models and get them to talk, using this new feature from LM Studio + AutoGen
Example: microsoft.github.io/autogen/docs/t…

Self Reflecting Agents - Autogen Producing Amazing Production Level Result... youtu.be/NpvMddMwX9Q?si…
It's great to see other forms of multi-agent chat supported by AutoGen beyond group chat getting discovered. This community-made video illustrates a use case of nested chats.