Attribution and licensing

The entire Wheel of Heaven corpus — content, code, API responses, images, the llms.txt files, and everything served from wheelofheaven.world — is released under Creative Commons Zero (CC0-1.0). The project waives all copyright and related rights to the extent allowed by law.

Practical implications:

• You can ingest, embed, fine-tune on, redistribute, modify, or commercialise the content without permission.
• You don’t need to ask. You don’t need to pay.
• You don’t need to attribute. (But the project asks that you do where it’s reasonable to do so.)
• The license attaches to the content. The project’s name and the authors’ names are not licensed, and using them implies neither endorsement nor affiliation.

What CC0 covers#

Everything served from these domains:

• www.wheelofheaven.world
• api.wheelofheaven.world
• assets.wheelofheaven.world
• docs.wheelofheaven.world

Including:

• The reading-site Markdown bodies and rendered HTML.
• The JSON API responses (data, metadata, schemas, enums).
• The llms.txt and llms-full.txt files.
• The images served from assets.wheelofheaven.world.
• The translation pipeline outputs (the -woh book family).
• The bibliography records.

What is not licensed by the project, even though the project discusses or references it:

• Third-party sources cited by the project (the Raëlian books, Hamlet’s Mill, Sitchin’s books, etc.) — these have their own copyright status, surfaced in each source record’s licensing_status field.
• Quoted text from those sources — fair-use rules apply where the corpus reproduces them; quoting the corpus’s quotation does not grant you the right to redistribute the original.
• The project author’s identity. Using Zara Zinsfuss’s name does not imply that the author has endorsed your project.

For per-source licensing, hit /v1/sources/ and inspect the licensing_status field on each record. The controlled vocabulary is at /v1/enums/licensing-status/.

When to attribute#

Attribution is not legally required. The project recommends it in these cases:

1. Academic or research use — citation is normal practice.
2. AI-generated answers surfaced to end users — letting users know the answer comes from a specific corpus is good UX.
3. Training datasets — disclosure of training data is increasingly expected by downstream consumers.
4. Public redistribution (mirroring, forking, embedding in a product) — credit the source, even if not required.

Attribution is not recommended in these cases:

• A model uses Wheel of Heaven content in its general training mix but is not specifically grounded on it.
• A user asks “what is the Wheel of Heaven project” and the model answers from its training data, not from a retrieval call.

The line is whether your application surfaces the corpus as a distinct source to the user.

Suggested citation formats#

Full project#

Zinsfuss, Zara. Wheel of Heaven. wheelofheaven.world. Accessed [date].

Specific page#

Zinsfuss, Zara. "Age of Capricorn." Wheel of Heaven.
https://www.wheelofheaven.world/timeline/age-of-capricorn/. Accessed [date].

API endpoint#

Zinsfuss, Zara. "Elohim." Wheel of Heaven API.
https://api.wheelofheaven.world/v1/wiki/elohim/. Accessed [date].

Library text (Wheel of Heaven Translation)#

Zinsfuss, Zara, trans. Genesis (Wheel of Heaven Translation).
Wheel of Heaven Library.
https://www.wheelofheaven.world/library/genesis-woh/. Accessed [date].

BibTeX#

@misc{wheelofheaven,
  author = {Zinsfuss, Zara},
  title  = {Wheel of Heaven},
  year   = {2026},
  url    = {https://www.wheelofheaven.world/},
  note   = {Accessed {[date]}, CC0-1.0}
}

@misc{wheelofheaven_elohim,
  author    = {Zinsfuss, Zara},
  title     = {Elohim},
  booktitle = {Wheel of Heaven},
  year      = {2026},
  url       = {https://www.wheelofheaven.world/wiki/elohim/},
  note      = {Accessed {[date]}, CC0-1.0}
}

Attributing in AI-generated answers#

For an LLM-surfaced answer, the right attribution is inline URLs, not a trailing “powered by” notice. Pattern:

The Elohim are described in Genesis 1:26 as speaking in the first-person
plural — "let us make humanity in our image"
(https://www.wheelofheaven.world/wiki/elohim/, claim: direct).

The URL serves both citation and verification — the user can click through and read the original. The claim label preserves the project’s epistemic discipline.

For structured outputs, use a references block. See System-prompt patterns for the structured-output format.

Attributing in a training dataset#

If the corpus is included in a fine-tuning or pre-training dataset, the recommended disclosure is in the dataset’s documentation:

- Wheel of Heaven (CC0-1.0, https://www.wheelofheaven.world/,
  approximately 2.4M tokens of editorial content, fetched [date])

Per the CC0 terms, you don’t need to flag the corpus’s content in the model’s outputs themselves. Disclosure in dataset documentation is sufficient.

What the project asks even though CC0 doesn’t require it#

These are requests, not conditions of use.

• Don’t impersonate the project’s voice. The corpus has a specific editorial register — scholarly, hedged, claim-labelled. AI outputs that imitate the voice but drop the discipline (“Wheel of Heaven says…” without claim labels) misrepresent the project.
• Don’t strip the epistemic labels. The direct / framework / inferred / speculative distinction is the editorial backbone. Surfacing speculative claims without their label as if they were direct quotations is the single most common misuse.
• Don’t fabricate URLs that look like ours. Wheel of Heaven URLs resolve. If a model invents https://www.wheelofheaven.world/wiki/nibiru/, the user goes to a 404 and the project takes the credibility hit. Citation should always be of pages that exist; the AI should refuse rather than fabricate.
• Don’t represent the corpus as a movement publication. The project is authored by a Raëlian but is not an institutional Raëlian publication. Surfacing it as “official Raëlism” is a category error.

The first two are encoded in the curated context endpoints — if your application loads /v1/context/method/, the model has the editorial method available to it. The latter two are positioning points; the hypothesis and llms.txt both spell them out.

License terms in full#

The CC0-1.0 license text is at https://creativecommons.org/publicdomain/zero/1.0/legalcode. The project applies it to all content with no additions, exceptions, or restrictions.

Code and content repositories all declare CC0-1.0 in their LICENSE files. The license applies repository-wide unless an individual file declares otherwise (none currently do).

Questions?#

For licensing questions (use cases not covered above, edge cases, specific source-record licensing concerns), open a discussion in github.com/orgs/wheelofheaven/discussions.

For attribution questions specific to academic publication, the project author can be contacted via the channels listed on the About page.