Jesse Liberty

I’ve been looking at a number of different ways to build Agents. I’ve settled on two and will be documenting what I learn as I go:

• Building from first principles based on my course on Agentics at Johns Hopkins
• Building using Microsoft’s new Agent Framework

The advantage of the first is that you understand the underlying mechanisms in more depth; the advantage of the second is that a lot of the plumbing is done for you and you become more productive more quickly.

I will do the .NET work in C#, and probably do all the other work in Python. See my blogpost on why Python.

I will, to a degree, be documenting what I learn as I learn it, without infringing on copyright, of course.

As promised, I’ll be posting slides and commentary from my recent user-group presentation on AI (Boston Code Camp). One of my first slides talked about the near-term evolution of AI, defined as either 1-2 years or 6 months, depending on who you believe.

I divided the slide into two parts: good news and bad. The good news is the promise of enormously increased productivity, which may well lead to a higher standard of living across the board. Further, as AI progresses, we may see accelerated breakthroughs in many fields, most notably medicine and genetics.

In a special videoCast, Lance McCarthy of Progress Software dives deep into MCP, not only explaining what it is for and how it works, but demonstrating, in code, how it is done.

MCP (Model, Context, Protocol) is an open standard that lets AI models (read copilot, ChatGPT, etc.) connect to external tools. These tools can be databases, code you write, code you connect to, other AIs, etc. etc. It is the universal API for Agents.

Here’s a link to the code.

In examining my HyperVideo project I decided that the Bunny video was a great placeholder, but what I really wanted was my own videos. What better place to start than the videos I have up on YouTube. How am I going to do that?

To get started, I asked CoPilot what it would take. CoPilot generated a plan of action. I took that and opened an issue on GitHub which will document what I have to do.

When working with CoPilot in Visual Studio, the single most important thing is context. Context tells CoPilot what it is working on, what it should already know, what language conventions to use, etc., etc. CoPilot wants to help, but its memory is not the best, and like the character in Memento, the trick is to write everything down and remind it all the time… hence context.

The best way to get ahead of this is to create a copilot-instructions.md file in your .github directory (which you may need to create). In this file, you’ll want to put anything that CoPilot should know right from the get-go and for every session.

But how do you know what to put in there?

This series on AI is going to take two paths:

• Writing prompts
• Writing apps that integrate AI (see Hypervideo)

If you are writing apps, you have a couple choices (well, more than a couple). One choice is to stay in the Azure/.NET world, in which case C# is just fine. The other (set of) choice(s) is to work in the rest of AI development, in which case Python is the language of choice.

I’m taking a course on agentics from Johns Hopkins University, and that course is in Python. It turns out that learning Python is pretty easy if you are a C# programmer, and it is even easier if you are a C# programmer with access to CoPilot and/or ChatGPT for Python.

I resisted learning Python all these years, but I’ve come to believe that if you are serious about AI then Python is in your future. So, might as well bite the bullet and get started now.

Every two years or so I publish a list of the utilities I find indispensable. It is interesting to me which ones last year after year (e.g., Evernote) and which fall off the list—often because the functionality I want has been absorbed into Microsoft software I’m already using.

Here is 2026’s list…