METHOD
Epistemic status
Every rendered page separates confidence from authorship.
A claim can be low-confidence and human-written, or high-confidence and AI-assisted. Those are different questions, so the site reports them independently.
Confidence
| Brain category | Belief | Meaning |
|---|---|---|
#shower-thought | 1/5 | Thought about briefly; low confidence. |
#guru | 1/5 | Highly speculative; be skeptical. |
#schroedinger-uncertain | 2/5 | Low confidence after some thought and learning. |
#socratic | 4/5 | Reachable through logic and Socratic questioning. |
#godel-uncertain | 4/5 | High confidence, but fundamentally uncertain. |
#lindy-proof | 4/5 | Supported by having survived over time. |
#floating-point-error | 4.99/5 | Very high confidence, short of absolute certainty. |
#unspecified means the private source did not record a confidence category. It is not a claim of neutrality or certainty.
Authorship and AI
- #non-biological / AI-generated: the private source says AI produced more than 75% of the text.
- #human-in-the-loop / AI-assisted: the private source says AI produced less than 50% of the text.
- Likely human: no AI marker is present and the source existed before public ChatGPT. This is evidence, not proof, because earlier models existed.
- Unknown provenance: no reliable authorship record exists. The site refuses to guess.
- Imported excerpts: quotations belong to the named author. Selection does not imply endorsement or completion.
- System-generated: the page is an index or visualization assembled from already-published material.
Important boundary
Mechanical Markdown or HTML generation does not make prose AI-generated. Likewise, AI-generated tags do not make the underlying note AI-authored. Only explicit provenance or defensible history changes an authorship label.