Insights — internal
Enter the password to continue.
Content And Journeys Audit

Content audit · read from the live database · 16 August 2026

Content And Journeys

Every card, journey, slide and author in the live database, audited for completeness, coverage and reach. The library is in better shape than the product's use of it.

Cards
697all embedded
Missing a hook
23333% of the library
Catalogue reached
35.6%248 ever opened
Journeys
1576 empty slots
Journey completions
0of 307 enrolments
Authors
5112 with no cards

A source library worth keeping: organism.earth

Reviewed 24 August. Recorded here so we can draw on it, and so the findings do not evaporate into chat.

What it is. The Library of Consciousness — a free, curated archive of several thousand works on consciousness, philosophy, systems and evolution. Essays, books, lectures, seminars, recorded discussions, television episodes. Roughly 1.5 million visitors all-time, ~1,500 a day. Navigation is Minds / Quotes / Search / RSS / About.

The author overlap with our library is substantial — Alan Watts appears repeatedly, alongside Carl Sagan, Joseph Campbell, Bertrand Russell, Isaac Asimov, Michael Levin, Joscha Bach, Donella Meadows and Teilhard de Chardin. Same well, and they publish the full original texts with clean provenance.

The critical caveat, before anyone treats this as a content pipeline. Organism.earth hosting a work does not grant Insights any right to reproduce it. Much of what is there remains in copyright, and their hosting arrangement is theirs, not ours.

Its value to us is as a provenance resource, not a content licence: a fast, reliable way to find and verify the original work behind a card — exact title, subtitle, year, document type. That is precisely what the 43 sourceless cards and the unverifiable quote slides need. Use it to check attribution, not to lift text.

Three things worth taking.

  • Their attribution discipline. Every item carries author, exact title, subtitle, year and document type — Essay, 1930; Lecture, 1969, 26:54; Book, 2008. It is a working model of the standard we do not currently meet.
  • Topic granularity. They run roughly 44 topics, each with a written definition and a document count — Metaphysics 310, Consciousness 305, Ontology 263, Eastern Philosophy 211, Emergence 159. We have eight topics for 697 cards. Eight buckets is too coarse to diversify against: when the recommender pads with nearest neighbours, “Mental” constrains nothing. Finer topics would give the deck something to be diverse along — which bears directly on the repetitiveness Marie reported.
  • Document type as a first-class field. Lecture, essay, book, seminar, interview. Each sets a different expectation and each is a retrieval dimension. We flatten everything into “card”, so a Rumi poem and a podcast argument arrive identically. That flatness is part of why the deck feels samey.

Where Insights is genuinely different, and why this is not a competitor to beat on depth. It is a library — you have to already know what you want and go looking. Nobody casually browses 310 metaphysics documents. Insights is a daily deck that learns you and builds a self-portrait from what you read. That difference is the moat.

But it does set a credibility benchmark. Someone who reads organism.earth and then meets an Insights card with an untraceable quote — or a misattributed one, as with the al-Bīrūnī card — will notice. That raises the stakes on quote verification rather than changing the plan.

How to use it, concretely.

  1. When verifying a quote, search their library for the author first — it gives the exact work and year faster than an open web search, and with better provenance.
  2. For the TED-licensed cards, check whether the same thinker made substantially the same argument in a work listed there. If so, that is a legitimate re-source — provided the card genuinely could have come from it.
  3. Treat their taxonomy as a reference when we eventually widen our eight topics. Not a copy — a reference.

The library is genuinely good

Worth saying first, because the problems below are all about use, not quality:

  • 697 cards, every one embedded — no gaps in the vector index, so the recommender can reach the whole catalogue.
  • Every card has a topic, a cover image, an author and slides. Zero exceptions on all four.
  • Average 7.2 slides per card, minimum 3, maximum 16. Not one card has 2 slides or fewer — there is no thin filler in here.
  • ~261 characters per slide. Substantial without being heavy.
  • 511 authors across 697 cards — roughly 1.4 cards per author. This is a genuinely broad expert base, not five people recycled.
  • Only 1 card lacks a reflection prompt, and only 2 authors have no cards.

A third of the library has no hook

On a swipe deck the hook is the card — it is the line someone decides on. 233 of 697 cards don't have one.

TopicCardsNo hookAuthorsNever served
Philosophical14766 45%13247
Psychological13954 39%11141
Physical5525 45%4415
Relational9436 38%7924
Science5716 28%495
Emotional6713 19%599
Spiritual7012 17%609
Mental6811 16%618

It is not evenly spread. Spiritual, Mental and Emotional are 16–19% incomplete. Philosophical and Physical are at 45%. Your two largest topics — Philosophical and Psychological, 286 cards between them — account for 120 of the 233 missing hooks.

The existing hook audit report flags 57 🟠 and 103 🟡 for rewriting. That is a different, smaller problem than 233 cards with no hook at all. Both are user-facing copy, so both are yours.

Two thirds of the library has never been seen

248 of 697 cards have ever been opened — 35.6%. 158 have never even been served. And only 23 cards in the entire catalogue have been opened five times or more.

So the practical library is a few dozen cards deep, not 697. You are paying to author, embed and store a catalogue that users are not reaching — and the reason is upstream: with 130 people ever activating and a median session of one card, there simply have not been enough reads to surface it.

This is not a content problem to fix with more content. Adding cards makes the reached fraction smaller. It is an argument for fixing activation first, then judging the library on what people actually meet.

The journeys are enrolled but not travelled

This is the sharpest finding in the audit.

307 enrolments. 274 have never taken a single step. Zero completions, ever.89% stalled

The furthest anyone has reached on any journey is position 7.9 of 20, on Love After the Fantasy. Every other path's leader is below position 4. Fourteen of fifteen paths have an average position of 0.0–0.4.

People choose a journey during onboarding — that step is mandatory — and then never come back to it. Which lines up exactly with the activation lockouts: they pick a path, hit the paywall, and cannot get back into the app.

JourneyEnrolledNever startedFurthestCardsEmpty slotsNo hook
unlearning-who-you-were-told-to-be32292.010101
healing-whats-underneath32291.01824
making-something-from-nothing28270.81553
understand-and-reset-your-nervous-system26231.61825
philosophy-for-when-life-doesnt-make-sense25231.01374
the-internet-broke-your-brain24220.41736
consciousness-and-the-mystery-of-being22190.81820
your-body-knows-before-you-do20180.81826
relationships-beyond-romance18154.01828
love-after-the-fantasy16117.91828
systems-thinking-building-a-better-world15142.01377
how-the-world-got-this-way15150.07131
body-health-foundations14121.01557
the-anatomy-of-desire1191.01376
women-who-rewrote-the-rules981.81375

Every journey is designed for 20 rungs and none is full

76 empty slots across 15 paths. The worst two are also revealing:

  • How The World Got This Way — 7 cards for 20 positions, missing 1, 3, 4, 5, 6, 8, 9, 10, 11, 13, 15, 17, 18. It has no position 1 at all; its first card sits at position 2.
  • Unlearning Who You Were Told To Be — your joint most-popular journey at 32 enrolments, and it is half empty: 10 cards, missing positions 6, 9, 10, 12, 14–19.

Not a bug, though — I checked. The path engine selects with .gt("position", current), so it takes the next card above where you are and skips gaps cleanly. A missing position 1 costs nothing functionally. The cost is editorial: a journey sold as a 20-step arc that delivers 7 rungs is a different product from the one described.

Consciousness And The Mystery Of Being is the model. 18 cards, 18 distinct authors, only 2 empty slots, and zero missing hooks — the only journey in the set that is editorially complete. Worth using as the reference for what "finished" means.

78% of cards belong to no journey

543 of 697 cards sit outside every path. Only 224 card-slots are used across all 15 journeys, and 154 distinct cards fill them.

That is not automatically wrong — the daily deck and the recommender draw from the whole catalogue, and journeys are meant to be curated spines rather than the whole library. But it does mean the journey feature touches under a quarter of what you have, while the 76 empty slots sit unfilled. The cards to fill those gaps may already exist.

910 quotes sit in quotation marks under a real person's name

This is the finding I'd act on before anything else on this page. It is not the copyright problem I described to you earlier — it is a sharper one, and it is not limited to the 43 cards I first pointed at.

First, a correction to what I told you. I said 43 cards had "no source and no legal defence at all." That was wrong. All 43 do carry a source — the name of the work — and for many it is a real, citable one: Flow, Attached, Give and Take, Mending Wall, On the Origin of Species, Determined, Being and Nothingness, Emergent Strategy. Naming the author and the work is the attribution the quotation exception asks for. A book has a title and no URL.

What was actually broken was the app. The reader only rendered its Source section when a URL was present, and CardDetailsItem never even carried the source title — the mapper read the URL and dropped the title on the floor. So the credit existed in the database and was shown to nobody. Fixed and building clean; ships with the next build.

The real issue is the quote slides. 40 of those 42 cards end on a slide that puts words in quotation marks and signs them with a real, mostly living person's name — Adam Grant, Sapolsky, Sabine Hossenfelder, Bettany Hughes, Banksy, Brian Eno. For those 40 there is no source anyone can check the words against.

I measured how closely each quote tracks the card's own hook, across all 627 cards that have both:

CardsWith a quote slideAvg similarity to the card's own hookNear-identical
Have a source URL5870.14428  (4.8%)
No source URL400.3078  (20%)

Where there was no real source, the "quote" is four times more likely to be the card's own hook in quotation marks. Bandelli's card hook reads "Purpose isn't a destination you arrive at—it's a frequency you tune into"; his quote slide reads the same sentence, in quotes, signed with his name.

I want to be careful here: similarity is not proof. A writer who reads a real quote and then writes a hook echoing it produces exactly this pattern, innocently. The direction can't be settled from the data. What the data does settle is that for these 40 there is nothing external to check against — and that they are the group where the pattern concentrates.

One is confirmed wrong. The card The Mathematics of not Knowing attributes to Abū Rayḥān al-Bīrūnī the line beginning "The seeker after truth is not one who studies the writings of the ancients…". That passage is Ibn al-Haytham's, from Aporias against Ptolemy (Sabra's translation) — one of the most-cited lines in the history of science, under the wrong name. The card has been served once, with 7 interactions.

One is confirmed right, and now properly sourceable. The Alan Watts line "The only way to make sense out of change is to plunge into it, move with it, and join the dance" is genuine — it is from The Wisdom of Insecurity (1951). The card currently credits it to "Alan Watts on Flow and Resistance". That can become a real citation.

Both took one search each. The remaining 38 are the same job, and it is mine, not Desktop's.

Correction — I scoped this too narrowly, twice. Having a source URL does not make a quote safe. The Sapolsky card cites Determined, a real book, and its quote slide reads:

"Show me a decision being made, and I will show you the chain of biology that produced it."

What Sapolsky actually wrote is "Show me a neuron (or brain) whose generation of a behavior is independent of the sum of its biological past, and for the purpose of this book, you've demonstrated free will." Same rhetorical shape, different words. It is a paraphrase inside quotation marks — close enough to look sourced, but not what he wrote.

That is a third failure mode, distinct from the other two, and it sits on a properly sourced card. So the job is not "verify the 40 sourceless ones" — it is verify the quote slides, full stop. The 40 are where it is worst and where it is unverifiable; they are not where it ends. 627 cards carry a quote slide.

I would not have found this by auditing sources, because the source was fine. Only reading the quote against the actual text finds it.

Why this outranks the licensing question. A licence problem is a dispute with an organisation about terms. A quote in quotation marks under a living person's name that they did not say is a false statement about that person, on a paid product whose entire premise is that the thinkers are real. It is also the one a single user can spot and screenshot.

Two decisions for you. Neither is mine to make — both are user-facing copy.

  • The al-Bīrūnī card. Reattributing the quote to Ibn al-Haytham leaves the card's author wrong; replacing the quote with a genuine al-Bīrūnī line is new copy. Which?
  • The other 39. I verify each one and bring you a table — confirmed / wrong / cannot verify — and you decide what happens to the ones that can't be confirmed. I'd suggest the unverifiable ones lose the quotation marks rather than the card, but that is a copy call.

What I would do, in order

  1. Fill the two worst journeys before the launchHow The World Got This Way (7 of 20) and Unlearning (10 of 20, joint most popular). Candidates likely already exist among the 543 unassigned cards; matching them is mechanical and I can propose a shortlist for your approval.
  2. Hooks for the cards people actually meet first. Not all 233 — start with the ones inside journeys and in the onboarding set. That is a few dozen lines of copy, not a few hundred.
  3. Do not add cards yet. 64% of the library has never been opened. More content makes that number worse, not better.
  4. Re-audit journeys after activation is fixed. Zero completions today tells you almost nothing about whether journeys work — nobody has been able to travel one. It is a measurement of the lockout, not of the feature.

Your feedback on this page