agricidaniel/fablesecondbrain — explained in plain English
Analysis updated 2026-05-18
Build a personal, source-cited knowledge vault tracking AI model behavior and agent workflows.
Create a separate client vault, add source material, and generate linked notes automatically.
Publish a sanitized public website version of research notes while keeping private captures out.
Audit a research vault for missing citations or unverified claims before calling it complete.
| agricidaniel/fablesecondbrain | 8thpark/geode | codeweb-dev/spark-ui | |
|---|---|---|---|
| Stars | 25 | 25 | 25 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | — |
| Complexity | 3/5 | 2/5 | — |
| Audience | researcher | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Python and Obsidian to browse the vault as intended, command line tool installs via pip.
This repository is a public knowledge base built as an Obsidian vault, organized around research into Claude Fable 5, Claude Mythos, Claude Code, and general practices for operating AI agents. The author is explicit that this project has no official connection to Anthropic and that everything in it comes from publicly available sources. The core idea is to turn scattered public information into linked, source-cited notes. Every note carries metadata, links to other notes, a citation back to its original source, and a confidence tag, so the whole vault forms one connected graph that can be browsed in Obsidian or on a companion website built with GitHub Pages. Alongside the notes, the project keeps a ledger of sources and a separate ledger of verified claims, so it stays clear where each piece of information came from and whether it has been checked. Audit scripts check the vault's completeness and consistency before it can be marked ready. The project describes itself as advisory only. It does not change any accounts, production systems, or live infrastructure, and it does not grant any rights to third-party material or make legal claims. Before anything is published, the README lays out rules such as keeping raw source captures and any credentials, tokens, or private data out of the public site, and using names like Anthropic, Claude, Fable, and Mythos only in a descriptive way, not as an endorsement. Beyond the note vault itself, the repository includes a Python package and command line tool for working with these vaults. This lets someone create a new named vault for a project or client, add new source material to it, run a synthesis step to generate notes and claims from that material, produce visual reports, and check what to do next. A separate demo vault is included so the tools can be tried out without touching the real research vault. The repository also ships a sanitized, public-safe version of the site under a site folder, deployed through a GitHub Pages workflow, which intentionally leaves out private notes, raw captures, and any account-specific details.
A public, source-cited Obsidian knowledge vault and toolkit for tracking research on Claude models and AI agent operating practices.
Mainly TypeScript. The stack also includes TypeScript, Python, Obsidian.
The README does not state a license for this project.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.