Alexandria

Building a library for human flourishing.

Live from the Cloudflare pipeline (talk-transcripts Worker → D1). Rebuilt from /api/graph, /api/protocols, /api/citations on every deploy — not a hand-maintained snapshot.

Pipeline progress

Live per-source counts, straight from D1, fetched fresh every time this page loads — not a build-time snapshot, so it's accurate right now regardless of when the site was last deployed. Phase 3 (protocol synthesis) and Phase 4 (independent research verification — PubMed + ClinicalTrials.gov) run unattended on a Cloudflare cron, not anything triggered from this session — see Citation verification → for the full claim-by-claim results.

Filter everything on this page

One filter state drives every chart and table below — click a bar, a cell, or type a search term, and the rest of the page narrows with it.

Mechanism graph — corroboration by concept

Distinct guests (by name) who raised each concept, across every transcribed episode. Click a bar to filter the co-occurrence heatmap, the specialist/testimonial split, and the drill-down table below to that concept alone.

Concept co-occurrence

Which concepts appear together in the same episode, and how often.

Click a cell above to drill into the actual conversations behind that pair (or a bar above to see everything that concept co-occurs with).

Guest × topic — who corroborates what

The concept graph above tells you a topic was corroborated by N guests; this tells you which guests, side by side, so you can see whether corroboration is broad (many different specialists, independently) or actually just one or two guests repeating themselves across episodes.

Click a cell to see exactly what that guest said about that topic.

Guest × guest — shared topics

Two guests who never appeared in the same episode can still be drawing the same connections independently — this is that overlap, computed from shared topics across the whole corpus, not just co-occurrence within one episode.

Explore any guest or topic in depth →

Specialist vs. testimonial

Correct speaker attribution matters more than raw claim count — a personal anecdote never gets to masquerade as a clinical claim, even from a credentialed guest. Faceted by citation flag: none_named is a bare assertion, anything else means the guest at least gestured at a source in conversation (see Citation verification for whether that source actually checks out).

Explore by author or topic → · Protocol candidates → · Citation verification → · Data sources →