Browse by what you care about.

Why I'm building a full AI system in public, piece by piece - and how to follow along (and deploy it yourself).

The foundation - standing up a Model Context Protocol server on a serverless Azure Function App.

Wiring the MCP server to the clients and data sources it talks to.

How data moves through the system, end to end.

Shaping how the system behaves and stays on task.

One server, every client - connecting the same MCP to the tools you already use.

A checkpoint - the architecture as it stands so far.

The brains - retrieval, vectors, and a local service API wired internal and external, in one resource group.

Letting the system reach out to the wider world, safely.

Tying it all together into an intelligent agent that helps manage Azure.

Take the whole thing, install it, and build your own apps on top.

Everyone says 'agent' like it's magic. It isn't; it's a loop. Here's the whole idea in plain English.

What the model actually sees; and why it seems to forget things mid-conversation.

How words and ideas become numbers a machine can actually compare.

Retrieve-then-generate; what it is, and when it beats fine-tuning.

How a text-only model reaches out and acts on the real world.

One plug for every tool; why everyone is suddenly shipping an MCP server.

How you measure an AI system before shipping it, so 'is it better?' stops being a vibe.

Tokens, latency, and dollars; the real math behind a single call.