Most AI content shows tools and APIs. This hub focuses on something slightly different: why the patterns exist, what problem they solve, where they break, and the engineering judgment that separates working systems from demos.
This hub is written for engineers who build applications with LLMs — working with RAG, MCP, agents, tools, workflows, and the control surfaces that keep them safe in production.
The focus is building with models, not building the models themselves — application and system engineering, not training or internals.
Newest
- AI Agents in Practice — Part 6: Building the Production Agent Loop
- AI Agents in Practice — Part 5: Workflow, Agent, or Single LLM Call — How to Decide
- AI Agents in Practice — Part 4: Five Agent Patterns and the Control Surfaces That Make Them Safe
Choose a Path
MCP in Practice — Read from the beginning (complete, 9 parts)
How AI applications connect to tools, data, and external systems — from first principles to local builds to production concerns.
You'll leave knowing: why connecting AI to systems is harder than it looks, what MCP actually standardizes, and how to build and harden a working MCP server.
Four waypoints through the series:
- Part 1 — Why connecting AI to real systems is still hard
- Part 5 — Build your first MCP server (and client)
- Part 6 — Your MCP server worked locally. What changes in production?
- Part 9 — From concepts to a hands-on example
RAG in Practice — Read from the beginning (complete, 8 parts)
How retrieval-augmented generation actually works, where it fails, and how to build and reason about it step by step.
You'll leave knowing: why RAG exists, what chunking and retrieval actually decide, how to build a working pipeline from scratch, and what breaks once it goes to production.
Four waypoints through the series:
- Part 1 — Why AI gets things wrong
- Part 3 — How RAG works: the complete pipeline
- Part 5 — Build a RAG system in practice
- Part 8 — RAG in production: what breaks after launch
AI Agents in Practice — Read from the beginning (in progress, 6 of 8 parts live)
What makes a system an agent, why demos break in production, and how to build agents that hold up — a control loop with tools, state, and boundaries.
You'll leave knowing: why the same model that aces a demo confidently does the wrong thing in production, what an agent actually is in engineering terms, how the loop runs turn by turn, the five patterns with the control surfaces that decide whether each one is safe to ship, how to choose the right architecture shape before writing a line of code, and how to build a production agent from the loop up — with the verify-before-commit gate that keeps it honest.
Live parts:
- Part 1 — The Demo Worked. Production Didn't.
- Part 2 — What Makes Something an Agent
- Part 3 — How the Control Loop Actually Works
- Part 4 — Five Agent Patterns and the Control Surfaces That Make Them Safe
- Part 5 — Workflow, Agent, or Single LLM Call — How to Decide
- Part 6 — Building the Production Agent Loop
More coming through summer 2026.
Where to Start
New here? → MCP Part 1, RAG Part 1, or Agents Part 1
Want to build something? → MCP Part 5 or RAG Part 5
Care about the decisions? → MCP Part 4 or RAG Part 6
Care about production? → MCP Part 6 or RAG Part 8
If this kind of practical AI writing is useful to you, this page is the easiest way to see what exists — and it has a home of its own at aiinpracticehub.com.
Top comments (2)
Some comments may only be visible to logged-in visitors. Sign in to view all comments.