AI Decoded
What the models are, how they are trained, and what they cannot do.
AI Decoded explains the field from the mechanism up: what machine learning does, what a neural network is computing, and how transformers and large language models actually work.
It covers training and fine-tuning, embeddings and retrieval, and the practical vocabulary — tokens, context windows, inference — that the surrounding discourse uses without defining.
It is equally concerned with limitations: what these systems fail at, why they produce confident errors, and which claims about them are marketing.
Not online yet. This one is written and waiting to be deployed. When it goes live it will be published on this domain and linked from here.
Also in the library
Azure Decoded
Technology
Microsoft's cloud, mapped service by service.
AWS Decoded
Technology
Amazon's cloud catalogue, translated into plain English.