The Learning Harness
An opinionated (but malleable) macOS app for learning a technical field properly.
How it works
What the harness must do
01Personalised learning
The harness asks what you know, then writes the lesson to fit: a self-contained page, a diagram in every part, one worked example throughout, the field's own terminology. A second AI pass reviews and corrects the page before you read it.
02Prerequisites come first
Every topic sits on one ordered path. The next topic is the smallest step from what you already know, and the harness teaches nothing before its prerequisites.
Two topics complete, one in progress, the rest ahead. The path is an order, not a timeline.
03Verification is king
A concept is not understood because you read it. It passes a quiz the same day, then an explain-back on a later day, after a night of sleep. A concept that fails the explain-back is taught again on the spot and explained again the next day.
04The Harness should be spaced repetition pilled
Once a concept passes, its cards enter a review queue ordered so that each day covers several concepts instead of many cards from one. The harness then measures retention from the review history, not from how well you feel it went.
05Organised learning across topics
The harness keeps each subject on a structured roadmap and balances the active roadmaps in one learning queue. You can study several topics at the same time without losing your place or neglecting one.
Inside the macOS app
Today keeps the day clear
Today shows the next step, due reviews, recent sessions, streaks, and current mastery in one place.
Lessons stay beside Chat
The lesson panel stays open while Chat runs placement, quizzes, and explain-backs. You can reply with text, speech, or images.
Roadmaps keep every topic in view
Roadmap shows what is done, what is next, and what is ahead. Topics lets you pause a subject without deleting its progress. Its existing reviews continue.
Reviews stay in the app
The built-in reviewer manages your daily cards. Anki desktop is optional, and you can connect it in Settings if you prefer it.
Guides
- How to learn properly with AI
Four rules separate learning with AI from being answered by it. Two randomized trials in 2025 show what each side costs.
- The best way to learn in 2026
The two techniques the evidence ranks highest, what people do instead, and a daily routine that follows the ranking.
- How to learn fast and remember what you learned
Speed is measured a week later. The method with the least re-learning is the fastest.
- Upskilling in 2026: how to learn a technical field while working
Thirty to forty-five minutes a day, one concept in prerequisite order, measured by what you can explain.
- What an AI tutor must do that a chat window does not
Six requirements from two trials with opposite results.
- Spaced repetition with AI
Cards made from what you were just taught, after it passed a test, into a review queue.
- The Feynman technique with AI
An explain-back graded against the lesson, the day after you read it.
The rules the harness runs, in order, are on the method page. Practical questions are in the FAQ.
The Learning Harness is in a closed beta and not yet available.