Skip to content

Library

The library/ is a growing collection of MemexLab knowledge assets — synthesized notes distilled from papers, reports, and books, each with provenance, key ideas, and atomic-note candidates. It is the durable output the system exists to produce: read something, turn it into a citable asset, let it compound.

It currently holds a paper (Attention Is All You Need), a strategic report (Benedict Evans, AI eats the world), and several books (Skunk Works, Grace Hopper, The Art of Doing Science and Engineering). The current, always-up-to-date list lives in the generated index: library/README.md.

Add an asset (it slots in cleanly)

  1. Drop a markdown file into library/papers/, library/reports/, or library/books/ with frontmatter:
---
type: source
title: "Title"
author: "Author"        # or authors: [A, B]
year: 2024
url: https://...         # canonical source — we link, never vendor the PDF
tags: [paper, ...]
---
  1. Body: a short summary, the key ideas, and [[atomic note]] candidates.
  2. Regenerate the index:
python3 scripts/build_library_index.py

The index groups assets by type and lists title / author(s) / year — no manual editing.

Papers vs. the worked example

  • The worked example demonstrates the pipeline (ingest → extract → frameworks → progress) on a fixed set of sources.
  • The library is the output — the durable, growing corpus those passes produce, one asset per source.

Papers and reports link to their canonical source (e.g. arXiv, the author's site) rather than bundling the PDF: the asset is the distilled knowledge, not a copy of the source.

← Engineering & design · Worked example