Why this exists
AI is not magic. It is also not simple.
There is a strange pressure around AI right now. People feel they should already know what MCP, agents, context windows and multi-agent systems mean. So we learn the vocabulary first, then quietly hope nobody asks us to explain the machinery.
AI Explained is for getting past that. It begins with the automation engineers already understand, then makes each new capability visible: what the model does, what the harness does, what policy permits, which tool creates an effect, and where the evidence lives.
How we work here
Plain English first
We introduce the mechanism without assuming the learner is already fluent in the jargon.
Build one evolving system
The Engineering Troubleshooting Workbench gives every concept somewhere concrete to live.
Keep boundaries visible
Model output is a proposal. Code, policy, tools, state and events retain their own responsibilities.
Treat sources honestly
Specifications, established engineering practice, research patterns and vendor implementations are not the same kind of claim.
Make failure part of learning
A lesson is incomplete if it cannot show what changes when an assumption breaks.
Built by Mohamed Ansari
Mohamed is an Engineering Manager working across AI Platform Engineering, SRE, Observability and DevOps. This guide is shaped by the questions that matter after the demo: What is this system actually doing? What can it touch? How does it fail? Can we observe, explain and improve it?
This is a living field guide.
Protocols and providers move quickly. If something is inaccurate, unclear or missing the part that makes it useful, please say so. The source and curriculum are open for review.
View the source on GitHub →