Understanding language AI
01
How language AI works
Word embeddings, next-token prediction, and what it means that language itself carries meaning.
02
The model landscape
Not all AI is the same. A framework for understanding what models can do and how they differ.
03
How models are trained
Pre-training, instruction tuning, alignment, and distillation. What the model learned and why it behaves the way it does.
04
Beyond the chat window
A chatbot is one interface. Models can classify, extract, score, transform, and generate in structured formats.
05
Where your AI runs
Public cloud, organisational infrastructure, or your own machine. Location shapes every decision.
Working with context
From ad hoc to systematic
08
Starting with chat: prompt design
Know what you want, sense what the model defaults to, steer towards your goal.
09
Is there such a thing as AI language?
Default chatbot output has recognisable patterns. Designed output does not.
10
From single tasks to repeatable workflows
The shift from asking the AI once to having a process that uses AI at specific steps.
11
Agents and agentic workflows
What agents actually are, what they are not, and why the public discourse gets this wrong.
The integrated generation environment
12
The augmented generation environment
An environment where your data is curated, indexed, and accessible, and you are the designer.
13
Designing your environment
What data goes in, how it is indexed, what agents have access to, and what rules govern them.
14
Measuring and adjusting
Evaluation is not a one-off. You measure, adjust, and measure again. The human is the quality system.
Build your own