Worked example: four sources, one pipeline¶
A complete, validating run of the MemexLab pipeline over four real, public sources — showing the three capabilities end to end with cited artifacts, and the learning loop closing its own coverage gaps.
No source text is reproduced — only paraphrased claims with provenance. The full vault passes
validate_vault.py. Browse it:examples/worked-example/.
Sources¶
| Source | Author | Year |
|---|---|---|
| As We May Think (the essay that named the memex) | Vannevar Bush | 1945 |
| Skunk Works | Ben R. Rich, Leo Janos | 1994 |
| Grace Hopper and the Invention of the Information Age | Kurt W. Beyer | 2009 |
| The Art of Doing Science and Engineering | Richard W. Hamming | 1997 |
The pass¶
| Step | Skill | Artifacts |
|---|---|---|
| 1. Ingest | memex-ingest |
source notes — one per book, with provenance |
| 2. Extract | memex-extract |
11 cited concept items |
| 3. Frameworks | memex-frameworks |
lensed syntheses (first-principles · second-order · inversion; cross-source) |
| 4. Progress | memex-progress |
coverage, gaps, learn-next |
What to notice¶
- Provenance everywhere — every concept carries a
source:and a[[wikilink]]back to it. - Frameworks turn facts into judgment — extracted facts run through mental-model lenses yield why these ideas matter for agent memory.
- The learning loop closes its own gaps — pass 1 (Bush only) flagged
problem-2andproblem-4as unserved; reading Hamming closes both, and coverage now spans all five latticework problems. - Cross-source synthesis — the compounding stack connects Hopper (abstraction) + Bush (maintenance) + Hamming (judgment) + Skunk Works (execution).
- It validates —
python3 scripts/validate_vault.py examples/worked-examplepasses.