Extract Wisdom

# Examples — extract_wisdom

Three sample input/output pairs showing how the pattern performs across input types.

---

## Example 1 — Podcast transcript

**INPUT**:
> Naval: "The thing about reading is that you have to do it 10,000 times before it starts paying off. Most people read three books and quit. The compounding only starts when you stop counting. The book you read at age 40 is read by the person who's already read 200 books — that's why it lands different. Same words, different reader."

**OUTPUT** (excerpt):

## SUMMARY
Naval Ravikant explains that reading compounds only after enough volume that the reader, not the book, becomes the variable.

## IDEAS
- Reading three books and quitting misses the compounding entirely.
- The 200th book teaches more than the third because the reader has changed.
- Most "good books" are good because of the reader's prior reading.
- Volume reading creates a pattern-recognition substrate that interprets new books.

## INSIGHTS
- Skill compounding requires staying past the boredom threshold most quit at.
- The same input produces different outputs based on the receiver's structure.

## QUOTES
- "The compounding only starts when you stop counting."
- "Same words, different reader."

(Pattern continues through HABITS, FACTS, REFERENCES, RECOMMENDATIONS — empty if not mentioned.)

---

## Example 2 — Academic paper abstract

**INPUT**:
> "We propose a novel attention mechanism, sparse attention, that reduces transformer compute from O(n^2) to O(n log n) by restricting each token to attend only to its k-nearest neighbors in a learned embedding space. We demonstrate equivalent performance to dense attention on language modeling benchmarks with 4x reduced wall-clock training time."

**OUTPUT** (excerpt):

## SUMMARY
The paper introduces sparse attention via k-nearest-neighbor token restriction, achieving 4x training speedup with equivalent language modeling performance.

## IDEAS
- Most attention computation is wasted on irrelevant token pairs.
- Learned embedding space can rank token relevance before attention is computed.
- Reducing complexity from O(n^2) to O(n log n) preserves benchmark quality.

## FACTS
- Sparse attention with k-nearest-neighbors yields 4x wall-clock training speedup.
- Dense attention scales O(n^2) with sequence length.

## REFERENCES
- (none in the abstract — full paper would list citations)

---

## Example 3 — Book chapter

**INPUT**:
> "From Chapter 3 of Atomic Habits: 'Every action you take is a vote for the type of person you wish to become. No single instance will transform your beliefs, but as the votes build up, so does the evidence of your new identity. This is one reason why meaningful change does not require radical change. Small habits can make a meaningful difference by providing evidence of a new identity.'"

**OUTPUT** (excerpt):

## SUMMARY
James Clear argues identity change emerges from accumulated small actions, each acting as a vote toward the desired self.

## IDEAS
- Every action is a vote for the identity you're becoming.
- Belief change does not require radical change — only accumulated evidence.
- Small habits compound into identity by repetition.

## QUOTES
- "Every action you take is a vote for the type of person you wish to become."
- "Small habits can make a meaningful difference by providing evidence of a new identity."

## RECOMMENDATIONS
- Design habits that match the identity you want, not the outcome you want.
- Count actions as votes, not as steps toward a goal.