Documentation

Study Sessions

How to get the most out of adversarial study.

Starting a Session

learnlock> /study
# or just press Enter
	
DUEL SESSION - 5 concepts
Type 'skip' to skip, 'q' to quit.
Double Enter submits your answer.

Session Flow

See the concept name, source, and current challenge
Type your explanation and press Enter twice to submit
LearnLock models your belief and probes the weakest claim
Low-confidence findings are shown, but they are down-weighted before scoring
After up to 3 rounds, you get a reveal and the next review date

Writing Good Explanations

BAD - Too Vague

"Backpropagation updates weights using gradients."

GOOD - Specific

"Backpropagation computes gradients via chain rule on the computational graph. Starting from loss, it calculates dL/dw for each weight by multiplying local gradients backward."

Using Images (OCR)

Answer with a photo of handwritten notes:

> /path/to/notes.jpg

Extracting text from image...
Extracted: "Backpropagation uses chain rule..."

Requires: pip install learnlock[ocr]

YouTube Review Links and /visual

For YouTube concepts, LearnLock now shows a transcript-linked timestamp after the reveal. If you want to inspect the actual frame, opt in manually.

Review at 12:42: https://youtube.com/watch?v=...&t=762
Type /visual to inspect the on-screen frame for this moment.

learnlock> /visual

Understanding Scores

||||| Perfect (5) - No holes found

||||. Great (4) - Minor gaps

|||.. Good (3) - Some gaps

|.... Needs Work (1) - Critical errors remain

Scores are intentionally coarse. LearnLock currently uses 5, 4, 3, and 1; low-confidence findings can be surfaced in the reveal without dragging you all the way down to a failing score.

Configuration