podcast-wiki
CLAUDE.md14 starsChanged 37 days agoMITView on GitHub
A Karpathy-style LLM wiki built from podcasts the user listens to on Spotify. Claude maintains wiki/; raw/ is immutable source text. Primary purpose: learning topics across episodes (concept pages). Secondary: actionable takeaways. Tools: uv run tools/fetch_transcript.py --pending lists raw files not yet in the wiki. /ingest-podcasts runs the whole pipeline. uv run tools/lint_wiki.py runs mechanical wiki checks; /lint-wiki runs the full lint (checks + judgment pass + fixes). /dream consolidates Claude's project memory from recent sessions (uv run tools/dream_replay.py replays…
The file
We split this CLAUDE.md into the rules it gives an agent
7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
Write everything in English.
- Distinctive: line 24
Every claim links its episode, and the person who made it when identifiable.
- Distinctive: line 79
Read the raw file.
- Distinctive: line 83
Add new claims to Key ideas with episode + person links.
- Very specific: line 88
Write the episode page (its raw: field must be the exact repo-relative raw path.
- Distinctive: line 89
Create or update the show page and the people pages.
- Very specific: line 91
Update wiki/index.md for every page created.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Podcast Wiki
A Karpathy-style LLM wiki built from podcasts the user listens to on Spotify. Claude maintains wiki/; raw/ is immutable source text. Primary purpose: learning topics across episodes (concept pages). Secondary: actionable takeaways.
Layout
raw/<show>/<date>-<slug>.md— local only (gitignored: third-party transcripts, repo is public). Source text with frontmatter (spotify_id, show, title, release_date, spotify_url, language, source, fetched). Never edit.wiki/concepts/<slug>.md— PRIMARY. One topic, synthesized across all episodes.wiki/hubs/<domain>.md— a domain (health, investing, …): overview + its concepts.wiki/episodes/<show-slug>--<episode-slug>.md— short source note per episode.wiki/people/<slug>.md— who they are, episodes, linked concepts.wiki/shows/<slug>.md— show description + ingested episodes.wiki/takeaways/to-try.md— checklist of actionable items.wiki/takeaways/recommendations.md— books / tools / people, grouped by type.wiki/index.md— every page, one line each, grouped by hub ("Unsorted" for concepts with no hub page yet).wiki/log.md— append-only ingest log.
Tools: uv run tools/fetch_transcript.py --pending lists raw files not yet in the wiki. /ingest-podcasts runs the whole pipeline. uv run tools/lint_wiki.py runs mechanical wiki checks; /lint-wiki runs the full lint (checks + judgment pass + fixes). /dream consolidates Claude's project memory from recent sessions (uv run tools/dream_replay.py replays them).
Conventions
- Write everything in English. Translate Hebrew sources during synthesis; proper names in common English spelling.
- Links: Obsidian wikilinks
[[concepts/zone-2-training]]. Slugs: lowercase-kebab-case English. - Every claim links its episode, and the person who made it when identifiable.
Page templates
Concept:
---
type: concept
hubs: [health]
sources: 4
updated: YYYY-MM-DD
---
# Zone 2 Training
**Summary:** 3–5 sentence synthesis of everything known so far.
## Key ideas
- Claim … ([[episodes/huberman--endurance]], [[people/peter-attia]])
## Disagreements & open questions
- Attia says 3–4h/week; Galpin argues 2h is enough … ([[episodes/...]], [[episodes/...]])
## Takeaways
- [ ] Try 4×45min zone-2 sessions/week ([[episodes/...]])
## Related
[[concepts/vo2-max]] · [[concepts/mitochondria]]Episode:
---
type: episode
show: <show name>
date: YYYY-MM-DD
guests: ["[[people/...]]"]
spotify_url: <url>
source: transcript|whisper|description
raw: raw/<show>/<file>.md
---
# <Episode title (English)>
5-line summary.
**Concepts:** [[concepts/a]] · [[concepts/b]]Person: frontmatter type: person; one-line bio; ## Appearances (episode links); ## Concepts (links).
Show: frontmatter type: show; short description; ## Episodes (links, newest first).
Hub: frontmatter type: hub; 2–4 sentence domain overview; ## Concepts (link + one line each).
Wiki update procedure
For each raw file from uv run tools/fetch_transcript.py --pending, one at a time:
- Read the raw file.
- Pick 3–10 concepts (1–3 if
source: description— thin source). For each, checkwiki/index.mdand grepwiki/concepts/for an existing page that fits; extend existing pages rather than creating near-duplicates. - For each concept page (new or existing):
- Rewrite the Summary to integrate the new material (don't append).
- Add new claims to Key ideas with episode + person links.
- A claim that conflicts with an existing one goes under Disagreements & open questions with both sources. Never silently overwrite an existing claim.
- Actionable items go in Takeaways.
- Set
hubs(1–2 domains), bumpsources, setupdated.
- Append each actionable item to
wiki/takeaways/to-try.mdunder the## <Hub>heading of the concept's first hub (## Unsortedif that hub has no page yet; same order aswiki/index.md) as- [ ] item — [[concepts/x]] · [[episodes/y]]. Add recommended books/tools/people towiki/takeaways/recommendations.mdunder the right heading, with the episode link. - Write the episode page (its
raw:field must be the exact repo-relative raw path; this marks the raw file as done). Copyspotify_urland IDs verbatim from the raw frontmatter; never retype them. Ifsource: description, add a line> Thin source: based on the episode description only. - Create or update the show page and the people pages.
- Hubs: for each domain with ≥5 concepts and no
wiki/hubs/<domain>.md, create it; otherwise update the existing hub's concept list. - Update
wiki/index.mdfor every page created. - Append to
wiki/log.md:## YYYY-MM-DD — <show>: <episode title>thensource: …,created: …,updated: ….
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

