The Roadmap /

Zero to AI engineer.

Most AI tutorials are a pile of tips with no order to them. This is the order. Ten stages, each one building on the last — starting with how a language model actually works, ending with agents, MCP and evals you can put in production.

The map

Ten stages, in order.

32 topics are published today and 36 more are queued. New tutorials land Monday, Wednesday and Thursday, with a hands-on project every Friday.

00

Orientation

What an AI engineer actually does, and how to set up a workspace you will keep using.

beginner 5 of 5 written
01

How LLMs Work

The mental model everything else is built on. Skipping this is why most people plateau.

beginner 3 of 8 written
How LLMs work, in plain English

Next-token prediction, transformers, and why a model has no memory between calls.

Read
Tokens and tokenisation

What a token really is, why cost and limits are measured in them, and how to count.

Read
The context window

The model's working memory: what fits, what falls out, and what that costs you.

Coming soon
Token usage and pricing

Reading a usage object, estimating spend, and cutting a bill without losing quality.

Coming soon
Why models hallucinate

Where confident wrong answers come from, and the habits that catch them early.

Coming soon
Temperature, top-p and sampling

The knobs that control randomness, and when changing them actually helps.

Coming soon
Open vs closed models

Hosted APIs, open weights, and how to choose without reading a benchmark table.

Coming soon
Reasoning and thinking modes

When extended thinking earns its cost, and when it is pure overhead.

Read
02

Prompt Engineering

Reliable prompting is the highest-leverage skill on this map. Everything downstream depends on it.

beginner 7 of 8 written
03

Context Engineering

Prompting is one message. Context engineering is managing everything the model can see.

intermediate 5 of 7 written
04

Embeddings & Vector Search

Turning meaning into numbers. The retrieval half of RAG starts here.

intermediate 0 of 5 written
What are embeddings?

Text as vectors, and why "similar meaning" becomes "close together".

Coming soon
Choosing an embedding model

Dimensions, cost and quality — the tradeoffs that actually matter.

Coming soon
Vector databases

Pinecone, Chroma, Qdrant, pgvector: what they do and when you need one.

Coming soon
Chunking strategies

How you split documents decides how good your retrieval can ever be.

Coming soon
Semantic search

Build search that understands intent instead of matching keywords.

Coming soon
05

RAG

Retrieval-augmented generation: give a model your data without retraining it.

intermediate 0 of 6 written
What is RAG?

The retrieve-then-generate loop, and the problem it exists to solve.

Coming soon
Building a RAG pipeline

Ingest, chunk, embed, index, retrieve, generate — end to end.

Coming soon
Retrieval quality

Re-ranking, hybrid search and filters, for when the right chunk never surfaces.

Coming soon
RAG vs fine-tuning

Knowledge or behaviour? The question that settles which one you need.

Coming soon
Grounding and citations

Make answers traceable to sources so hallucinations become visible.

Coming soon
Evaluating a RAG system

Measure retrieval and generation separately, or you will debug blind.

Coming soon
06

Agents

Models that choose their own next step. The biggest leap in capability — and in failure modes.

advanced 3 of 8 written
What is an AI agent?

The loop that separates an agent from a very good prompt.

Coming soon
Tools and function calling

Give a model real capabilities, and define them so it uses them correctly.

Coming soon
Building your first agent

A working loop from scratch: tools, results, and knowing when to stop.

Coming soon
Agent memory and state

What an agent carries between steps, and where to keep it.

Coming soon
Multi-agent systems

Delegation, sub-agents and orchestration — plus when one agent is plenty.

Read
Agent frameworks

Claude Agent SDK, LangChain, LlamaIndex: what each buys you.

Coming soon
Hooks and automation

Fire your own code at defined points in an agent's lifecycle.

Read
Multi-model workflows

Route each step to the model that handles it best.

Read
07

MCP

The Model Context Protocol: one standard way to plug tools and data into any model.

advanced 7 of 8 written
08

Fine-Tuning

Changing the model itself. Powerful, frequently unnecessary, and worth understanding either way.

advanced 0 of 5 written
When to fine-tune

The honest checklist — most problems are prompting or retrieval problems.

Coming soon
Preparing training data

Dataset size, format and quality, which decide the whole outcome.

Coming soon
Running a fine-tune

Walk one end to end and read what the training metrics are telling you.

Coming soon
LoRA and efficient tuning

Adapt a model without retraining all of it.

Coming soon
Evaluating a tuned model

Prove it beat the base model instead of assuming it did.

Coming soon
09

Evaluation & Production

The difference between a demo and something you can put in front of real users.

advanced 2 of 8 written
Why evals matter

Without a scoreboard, every prompt change is a guess.

Coming soon
Building an eval set

Real cases, expected outputs, and a script that runs in one command.

Coming soon
Detecting hallucinations

Automated checks that catch confident nonsense before your users do.

Coming soon
LLM-as-judge

Use a model to grade a model — and know where that breaks down.

Coming soon
Tracing and observability

See every call, token and tool result when something goes wrong in prod.

Coming soon
Cost and latency

Caching, batching and model choice, measured per completed task.

Coming soon
AI code review

Turn a model into a reliable second reviewer on your own work.

Read
Automated tests with AI

Generate a test suite that actually guards against regressions.

Read
Keep going