Project

About

The best ideas you encounter online disappear into the scroll. You bookmark a tweet, star a repo, save a thread — and never see it again.

Knowledge Engine changes that. Every piece of content you save gets broken into atomic insights — specific, actionable takeaways that stand on their own. Not summaries. Not tags. Real knowledge you can act on.

Those insights are routed into themed topics where related ideas from completely different sources sit next to each other. Patterns emerge that no single tweet, meeting, or conversation could reveal alone.

A personal research engine that remembers everything you found interesting — and tells you why it matters.

13

Topics

199

Sources

147

Voices

How it works

Architecture

Knowledge Engine is a repo-as-database system. No server, no build step, no vector store — just markdown files that an AI agent reads and writes. The repository is the entire persistence layer, browsable in Obsidian and version-controlled with Git.

Five-Stage Pipeline

/ingest

Raw content in, source files with extracted insights out

/synthesize

Route insights to themed topic files by pattern

/consolidate

Surface cross-cutting themes across all topics

/ask

Query your knowledge, get cited answers

/plan

Turn rough ideas into structured action with context

Tweet ingestion runs automatically on a nightly schedule. A Node.js tool fetches bookmarks from X via the Bird CLI, enriches each tweet with Claude API for insight extraction and topic matching, and writes source files directly to the repository.

This site is built with Astro and Tailwind CSS, statically generated at build time. In development it reads from the local filesystem; in production it fetches from the GitHub API. Deployed to Vercel on every push.

MIT Licensed

Open Source

The repo, the skills, and the ingestion pipeline are all open. Fork it, extend it, make it yours.

View on GitHub ↗