- Screened
- In the cut
- Culled
Type a topic.
Skip the bad lessons.
It writes the syllabus, then a screening agent works through YouTube chapter by chapter and keeps the one lesson worth your hour. Everything else stays on the bench.
The bench is what you
never have to sit through.
For every chapter it drafts a query aimed at a lesson rather than a trailer, then throws away everything that query turned up except the one result worth an hour of your evening.
Move the line
yourself.
Drag it left and the bench fills back up with everything the agent threw away. Drag it right and you are left with 13 lessons — which is the course.
Sets how deep into each chapter’s results the agent is allowed to keep looking.
The cut
Sixty-three seconds, real screen, nothing faked. The prompt is Typescript, the units are Basics and Advanced, and everything after that is the generator working — the outline, then every chapter, then a lesson video sitting in each one.
This deployment runs the recording only. Generating a course needs OpenAI and YouTube keys, which are not attached here — run it with your own.
What happens between typing a topic and having a course
- 01In the cut
You write the slate
In the cutA title, then as many units as you want to cover — Basics, Advanced, whatever you decide to call them. You draw the outline; nothing guesses at what you meant by the topic.
- 02In the cut
Every chapter gets a search of its own
In the cutThe syllabus is broken into chapters, and each one gets a query written to surface a lesson rather than a trailer. That query runs against YouTube and the highest-ranked teachable result takes the slot.
- 03In post
The agent reads the room
In postRank alone is a weak signal. A market-intelligence pass will score the whole shortlist for a chapter rather than trusting first place, and the lesson that survives all four reads is the one that ends up in your course.
- Rank position
- where the video lands for the chapter's query
- Watch signals
- views set against how long the video runs
- Top comments
- whether the people who finished it say it taught them
- Captions present
- no transcript means no summary and no questions
- 04In the cut
Transcript in, summary out
In the cutThe winning video's transcript is pulled and condensed, so you can read a chapter in thirty seconds and decide whether to spend the hour on it.
- 05In post
Then it asks you the questions
In postThe same transcript will drive concept checks and interview questions per chapter, aimed at the role you are preparing for — so finishing a chapter means answering for it, not just watching it.
Steps marked in post are being built and are not in the recording above. They are listed here because they are the next two things to ship, not because they already work.
One word in. Thirteen lessons out.
This is the syllabus the generator produced in the recording, transcribed exactly. Four units, thirteen chapters, a video and a summary behind every one of them.
- units
- 4units
- chapters
- 13chapters
- prompt
- 1prompt
- Unit 01
Introduction to TypeScript
- What is TypeScript?
- Setting up Development Environment
- Unit 02
TypeScript Basics
- Variable Declarations
- Data Types
- Functions
- Interfaces
- Unit 03
Advanced Topics
- Classes
- Generics
- Modules
- Decorators
- Unit 04
TypeScript Tools and Libraries
- TypeScript and Angular
- TypeScript and React
- TypeScript and Node.js
What it is built on
- Next.js 14
- App Router, generation runs server-side
- OpenAI
- syllabus, chapter queries, summaries
- YouTube Data API
- one lookup per chapter, throttled
- Prisma + PostgreSQL
- courses, units, chapters
- NextAuth
- Google sign-in
- Stripe
- free credits, optional Pro
Chapter lookups are throttled rather than fired in parallel: the YouTube Data API charges 100 units per search against a default 10,000-a-day quota, which works out at roughly a hundred chapters a day before it starts returning 403s.