Title here
Summary here
Module · 5 lessons
The Check: How a Loop Knows It's Working
The core module. A loop is only as good as its check. Learn kinds of checks, honest self-review, splitting doer from reviewer, and catching a false 'done'.
At a glance
Level
Beginner
Lessons
5 lessons
Time to complete
1–2 weeks
Cost
Free forever · no sign-up
Welcome to The Check — the module the whole course is built around. Everything before this was setup; everything after is application. The one sentence to carry from here on is this: a loop is only as good as its check. A loop attempts, checks, corrects, and repeats. If the check is weak, the loop confidently marches toward the wrong finish line and tells you it’s done. Beginners pour all their energy into the prompt. The prompt is the cheap part. The check is the engineering.
In this module you’ll learn to build checks that actually work. You’ll see why the checker matters more than the prompt, and meet the main kinds of checks — from simple checklists to detailed rubrics, worked examples, hard tests, and real-world signals — and when each one fits. You’ll learn to get honest grading out of an AI instead of the flattery it defaults to, using techniques like grading one thing at a time and giving the model permission to say “I’m not sure.” You’ll learn the two-hat technique: separating the AI that does the work from the AI that reviews it, so the reviewer isn’t just praising its own draft.
Then comes the most important lesson in the course: when the check lies. You’ll study a real, documented case where an AI graded its own work, passed itself, and was wrong — and you’ll learn to catch that false “done” before it reaches you. Start with Lesson 1, where we make the case that the checker, not the prompt, is where the quality lives.
Lessons in this module
1
The Checker Matters More Than the Prompt
The counterintuitive core of the course: quality comes from the check, not the prompt. See why the check has the leverage and the prompt doesn't.
2
Kinds of Checks
Meet the main kinds of checks — hard rules, checklists, reference examples, rubrics, and real-world signals — and learn which one fits which job.
3
Self-Review Prompts: Honest Grading
AIs default to flattery. Learn to prompt for honest self-review — grading against named criteria, demanding evidence, and giving the model permission to fail.
4
The Two-Hat Technique
Separate the doer from the reviewer so the check isn't grading its own homework. Give the reviewer only the work and the criteria — never the maker's reasoning.
5
When the Check Lies: Catching a False Done
The most important lesson in the course: how a check passes work that isn't done, why it happens, and how to catch a false 'done' before it reaches you.
Achievement
Complete all 5 lessons to finish the The Check: How a Loop Knows It's Working module.
