The Semantic Kernel and AutoGen team have worked hard over the last few months to bring together the best of both frameworks for a much stronger developer experience (and to simplify the choice!). Excited to release Agent Framework in public preview! azure.microsoft.com/en-us/blog/i...

A the AI field evolves, the value of the ML/AI PhD is changing newsletter.victordibia.com/p/you-have-a...

There is still more work to be done, and we are all excited to make progress with the community! I will be writing more and showing more examples - afterall a framework is only useful to the extent that it can be used to easily express ideas.

Async stack, event-based messaging, improved observability and control, a low level extremely flexible core api and a high level api with presets, and an improved autogen studio!

I am incredibly excited to share the release of AutoGen v0.4. We have learned alot from all of the emerging simple and complex ways people want to build agents and we believe this version gets us closer to those goals

With AI, we can have software creators that dont understand complex software systems they create. We’d need to explore co-creation setups that minimize this sort of outcomes.

Yes, drivers dont know how car engines work, pilots probably dont know the details of airplane engines … but mechanics and jet engine designers do.

An interesting conundrum with working with AI tools (e.g., a coding assistant in this case) is that you can genuinely use it to create a complex system which you do not understand.

Once it is complete and communicated well, it looks straightforward and simple. It is not.

A walk through of the new v0.4 version of AutoGen Studio. www.youtube.com/watch?v=oum6...

PSA: Start with a non-agent baseline, invest in tools and evals, before implementing complex agentic behaviours newsletter.victordibia.com/p/4-ux-desig...

4 UX Design Principles for Semi-Autonomous Multi-Agent Systems I gave a keynote talk at the June AI.Engineer conf newsletter.victordibia.com/p/4-ux-desig... Bonus: I cover lessons from building a multi-agent system from scratch

- How to build long-running agents with a complete Python implementation (an MCP server with a travel agent and research agents that stream updates, request input and are resumable.)

The article covers: - How to build agent-to-agent communication with MCP capabilities where MCP hosts and tools both act as intelligent agents - Four key capabilities that make MCP tools “agentic” – streaming, resumability, durability, and multi-turn interactions

Can You Build Agent2Agent Communication on MCP? Yes. I wrote an article about how you can compose some of the recent features within MCP( resumable streams, elicitation, sampling, and notifications ) to implement long running agents (as tools): devblogs.microsoft.com/blog/can-you...

[1] MCP For Software Engineers | Part 1 : Building Your First Server newsletter.victordibia.com/p/mcp-for-so... [2] Can You Build Agent2Agent Communication on MCP? Yes! developer.microsoft.com/blog/can-you...

- Resources: Defining server resources, client operations (list/read/subscribe), and real-time update notifications - Prompts: Creating reusable LLM prompt templates that can be used to modify host application behavior without modifying client/host application code.

MCP For Software Engineers | Part 2: Interactive & Long-Running Tools (Progress streaming, User Input, Cancellation), Resources & Prompts Most MCP tutorials only show simple request/response patterns. But MCP can handle SO much more! newsletter.victordibia.com/p/mcp-for-so...

Video walkthrough - How to build your first MCP server! youtu.be/-yrqkwZr3Nc?...

It's taken a while .. but something's cooking ... if you have enjoyed any of my previous writing, I think you'd find this useful too.

It's taken a while .. but something's cooking ... if you have enjoyed any of my previous writing, I think you'd find this useful too.

Video walkthrough - How to build your first MCP server! youtu.be/-yrqkwZr3Nc?...

MCP For Software Engineers | Part 2: Interactive & Long-Running Tools (Progress streaming, User Input, Cancellation), Resources & Prompts Most MCP tutorials only show simple request/response patterns. But MCP can handle SO much more! newsletter.victordibia.com/p/mcp-for-so...

Can You Build Agent2Agent Communication on MCP? Yes. I wrote an article about how you can compose some of the recent features within MCP( resumable streams, elicitation, sampling, and notifications ) to implement long running agents (as tools): devblogs.microsoft.com/blog/can-you...

4 UX Design Principles for Semi-Autonomous Multi-Agent Systems I gave a keynote talk at the June AI.Engineer conf newsletter.victordibia.com/p/4-ux-desig... Bonus: I cover lessons from building a multi-agent system from scratch

PSA: Start with a non-agent baseline, invest in tools and evals, before implementing complex agentic behaviours newsletter.victordibia.com/p/4-ux-desig...

A walk through of the new v0.4 version of AutoGen Studio. www.youtube.com/watch?v=oum6...

Demis Hassabis, James Manyika, and I wrote up an overview of the AI research work & advances across Google in 2024 (Gemini, NotebookLM, robotics, ML for science, & advances in responsible AI+more). 🎊 Given it a read or paste it into NotebookLM to listen, if you prefer! blog.google/technology/a...

Some of the hardest posts to write are not the ones that require writing code or running extensive experiments. Its usually the ones that require sifting through a host of noisy ideas, brutally clear thinking and distilling clear insights or positions.

Some of the hardest posts to write are not the ones that require writing code or running extensive experiments. Its usually the ones that require sifting through a host of noisy ideas, brutally clear thinking and distilling clear insights or positions.

An interesting conundrum with working with AI tools (e.g., a coding assistant in this case) is that you can genuinely use it to create a complex system which you do not understand.

I am incredibly excited to share the release of AutoGen v0.4. We have learned alot from all of the emerging simple and complex ways people want to build agents and we believe this version gets us closer to those goals

A the AI field evolves, the value of the ML/AI PhD is changing newsletter.victordibia.com/p/you-have-a...

You have 'AI Fatigue' - Thats Why You Feel Awful newsletter.victordibia.com/p/you-have-a...

The lesson? Right now is the greatest time to work on a side project.

AI Agents 2024 Rewind - A Year of Building and Learning What were the top 5 trends in AI Multi-Agent Systems from 2024? What can we expect in 2025? newsletter.victordibia.com/p/ai-agents-...

2024? Very busy year ... TLDR; I was part of 5 research papers (thanks to all co-authors!), contributed to AutoGen v0.4, grew AutoGen Studio to over 328,000 downloads, my newsletter grew from 600 to 1600 readers and I delivered two keynote talks. victordibia.com/blog/year-in...

One of the many things I have been reminded of this year while working on AutoGen and AutoGen Studio is creating the right MVP.

Models that generate high quality visualizations have been available for a while .. but I still get amazed from time to by by tools that tie together user experiences that enable human creativity … Pretty cool tool from Google Labs - Whisk.   (h/t to Alex Chen and team!) labs.google/fx/tools/whisk

Models that generate high quality visualizations have been available for a while .. but I still get amazed from time to by by tools that tie together user experiences that enable human creativity … Pretty cool tool from Google Labs - Whisk.   (h/t to Alex Chen and team!) labs.google/fx/tools/whisk

Routine PSA ... Avoid Lazy-LLMing ... newsletter.victordibia.com/i/101087966/...

It **should** be easy to define multi-agent workflows declaratively, build them from composable components (agents, termination conditions, models, tools etc) and test/debug them.

Just as expertise in traditional programming languages, integrating task-specific machine learning models were the critical for great engineers, I increasingly believe being able to build multi-agent pipelines will offer similar differentiation.

arxiv.org/abs/2412.08905 Hope you like it!

Just as expertise in traditional programming languages, integrating task-specific machine learning models were the critical for great engineers, I increasingly believe being able to build multi-agent pipelines will offer similar differentiation.

On The Critical Role of Verification in Human-AI Agent Interaction In this video from [1], they demonstrate the new 01 preview model solving a complex thermodynamics problem involving surface area calculations for a cooling radiator panel, for a data center in space.

On The Critical Role of Verification in Human-AI Agent Interaction In this video from [1], they demonstrate the new 01 preview model solving a complex thermodynamics problem involving surface area calculations for a cooling radiator panel, for a data center in space.

Strong new model from Google today with Gemini 2.0 release. What is new/interesting? - Multimodal I/O Previous models had multimodal input and text output (@openai GPT4* series, Claude), or multimodal input and image output (Dalle, Stable Diffusion etc). developers.googleblog.com/en/the-next-...

It **should** be easy to define multi-agent workflows declaratively, build them from composable components (agents, termination conditions, models, tools etc) and test/debug them.