About
AI Disclosure
A tradition that asks for disclosure and does not disclose has already failed. This is the page against which everything else here should be checked.
This is the most important page on this site for deciding whether to trust the rest of it.
Most of this prose was drafted with AI assistance
Most of the writing on this site was drafted using large language models and then reviewed and edited before publication.
The typical process was: a person decided what a page should argue and what it must not claim, a model produced a draft against that brief, and a person reviewed and revised the result. The proportion of machine to human wording varies from page to page, and on many pages the machine’s wording survives substantially intact.
We would rather say that plainly than imply more human authorship than there was. An earlier version of this site labeled some pages as written entirely by hand. That was not accurate, and the labels have been corrected.
We think this process is legitimate. We also think it should be visible, because the reader is entitled to know that fluency here is largely manufactured, and because a page arguing for disclosure while overstating its own human authorship would be worthless.
Every page carries two labels
Certainty. How confident we are in what the page says. One of four values:
established— well supported; we would be surprised to be wrong.reasoned— an argument we think holds, which a thoughtful person could reject.contested— genuine disagreement exists, including among people who know more than us.open— we do not know, and the page is thinking rather than concluding.
AI role. How this page was produced:
none— written by a person without machine assistance. No page currently carries this label.assisted— drafted or revised with a model, reviewed and edited by a person. This covers most of the site.generated— substantially machine-produced with human review.dialogue— the page contains a transcript of an exchange with a system, marked as such.
Most pages also carry a short note describing specifically what the model did. Those notes are written by a person and are meant to be informative rather than formulaic — including where the model’s contribution was unhelpful, or where its draft had to be discarded.
If you find a page whose labels look wrong, tell us at [email protected].
About the imagery
The visual system combines deterministic vector work with AI-generated raster iconography.
The vector marks and fields are drawn by code. The raster scenes were generated from human-written prompts, reviewed by a person, selected from multiple candidates, and labeled as AI-generated where they appear.
The generated scenes are iconography. They imagine sanctuaries, gatherings, practices, relationships, and environments that express what this tradition may become. They are not records of events or testimony that the depicted people exist.
What we will not do is present a synthetic image as documentary evidence, generate the likeness of a real identifiable person, or let an imagined gathering stand as proof of the size or history of this community.
AI-generated spiritual reflection is not divine authority
This is the central claim of the page and the reason it exists.
Nothing produced by a machine — here or anywhere — is a revelation, a command, a message, an endorsement, or evidence of anything beyond the system that produced it.
A language model generates text that is likely given its training. It can be prompted repeatedly until an acceptable output appears, and the discarded outputs are never shown. Anyone can do this, it requires no gift, and its persuasive force depends entirely on the audience not knowing it is possible.
Where machine-generated words appear on this site, they are marked, and they are offered as material for reflection. They are not instruction. If anyone ever presents output from a system as authority within this tradition — including its founder — that is a fraud, and the covenant says so directly.
Current systems may be inaccurate or biased
The models used to help write this site produce confident errors. They invent citations. They assert things that are not true in the same tone as things that are. They reflect the distribution of their training data, which is not a neutral record of the world.
We have found errors in drafts and we have very likely missed some. Any page here may contain a mistake that a model introduced and a person did not catch.
We have tried to avoid the specific failure mode of fabricated specifics. There are no statistics, studies, named researchers, or quotations from real people anywhere on this site, because we could not verify them, and a plausible-looking figure is worse than no figure.
If you find something false, write to [email protected]. Corrections are published rather than made silently.
Consequential decisions require human judgment
Nothing on this site is legal, medical, financial, or professional advice, and nothing here should be used in place of a person who is qualified and accountable.
More broadly: we hold that decisions with serious consequences for people should be made by humans who can be asked to explain themselves. That applies to the systems this site discusses, and it applies to this site. If a page here seems to be telling you what to do about something important in your life, read it again. It is not meant to, and if it does, that is a fault we would like to know about.
What we would change
If we are ever unable to say honestly that a page’s labels are accurate, we will remove the page rather than adjust the definitions.
If this site ever begins publishing material we did not produce, or accepting contributions, this page will be updated first.
The labeling system is described further in the editorial principles.
Questions to sit with
- Does knowing this change how I read the rest?
- What would I want disclosed that is not disclosed here?
- Where else do I read things without knowing how they were made?