Aug 2026 edition · reissued every month
State of Models — Aug 2026
The frontier model landscape this month, and which one to actually use for each job.
Rewritten from scratch each month and dated on the cover, so you always know how old the advice is — and every figure carries the source it came from.
Updated
- lessons
- 17
- new edition
- Monthly
- access
- Lifetime
Then go deeper: the AI path
Knowing which model to pick is the start. This four-course path takes you from knowing nothing to pricing a feature, wiring an assistant into your systems, and building an agent that survives contact with real work.
Each level assumes the one before it and nothing else. Take them in order, or start wherever you already are — every course says plainly who it is for.
- courses
- 4
- lessons
- 125
- access
- Lifetime
AI Basics
beginnerWhat an AI assistant is actually good for, where it will let you down, and how to use it on real work without getting burned.
You use a computer for work — email, documents, maybe a ticket queue · 41 lessons · updated Aug 2026
$14.99View course →AI and LLM Foundations
beginnerThe working vocabulary and mental models nobody can bluff you on.
No maths, no programming — curiosity is enough · 26 lessons · updated Aug 2026
$19.99$29.99View course →MCP in Practice
intermediateThe Model Context Protocol, from the handshake to a server your team actually uses.
Comfortable with HTTP, JSON and a language that can run a server · 27 lessons · updated Aug 2026
$24.99$34.99View course →LLM Workflows
intermediateLoops, tools, skills and subagents — and knowing which one the problem needs.
Comfortable reading code in any language · 31 lessons · updated Aug 2026
$39.99View course →
Why these and not a blog post
Real numbers you can check
Costs are worked through with current model prices, line by line, down to the monthly bill. Every calculation is shown, so you can check it rather than trust it.
You do the work
Every section ends with exercises and worked answers — cost estimates to compute, claims to diagnose, broken agent loops to find the bug in.
Code that runs
The later levels ship complete files, not fragments: a working agent loop, a server you can point a client at, an evaluation harness that catches regressions.
Nothing invented
No made-up benchmarks, no fabricated adoption statistics, no measurements nobody took. Where a figure is illustrative, it says so.
Also on the platform
Outside the AI track, for when you need something else.
Data Modelling for MongoDB
Need to Know HQ
No ratings yet
7 lessons
Updated Jul 2026
Production Web Services in Rust
Need to Know HQ
No ratings yet
12 lessons
Updated Jul 2026
Skills, Subagents and Tools
Need to Know HQ
No ratings yet
28 lessons
Updated Aug 2026
Buy it once, keep it
A course you buy is yours for good — no rental, nothing to renew.
Progress that sticks
Pick up exactly where you left off, on any device.
Reviews you can trust
Only verified students who bought the course can rate it.
Start where you are
Level I assumes no technical background at all. Every level has free preview lessons, so you can read before you buy.