Stay Learning

One prompt in. A whole course out — curriculum, lessons, exercises, quizzes and a capstone, in one house voice. Every file checked by a validator that blocks whatever it cannot verify.

/new-course
topic     Linux & Kubernetes networking with eBPF
audience  platform engineers, comfortable with Linux

 audience · curriculum · outcomes · assessment
 5 lesson plans
 5 lessons · 5 exercise sets · 5 quizzes · capstone

validator: 0 failures

Not a one-shot prompt

Four properties separate a generated course from a plausible-looking pile of Markdown.

Research-backed design

Backward design, Bloom's Taxonomy, cognitive load theory, and retrieval practice — encoded as machine-enforced constraints, not optional suggestions.

Mechanical verification

A deterministic Python validator runs after every file write. It checks structural integrity, Bloom verb correctness, outcome coverage, and terminology consistency.

Parallel generation

Sequential planning stages ensure quality. Once all plans are current, lesson content, exercises, quizzes, and the capstone run concurrently without conflicts.

Web viewer included

A built-in Astro frontend reads the generated course files and renders them as a browsable learning experience with progress tracking, exercises, and quizzes.

How it actually works

The architecture, the parallelism, the validator, and the instructional design behind the generated material.

Read the technical deep-dives →About the project