Claim strength should match evidence
A content-heavy site can accidentally turn speculative language into confident claims. The evidence ladder prevents that. Each public claim should be labeled by the level of support behind it.
Ladder
| Level | Allowed wording | Example |
|---|---|---|
| Demonstrated | "does" or "has shown" | a tested build step passes validation |
| Established pattern | "commonly used" | canary release and rollback |
| Emerging method | "early work suggests" | evolutionary model merging |
| Conceptual framework | "this site defines" | model breeding umbrella term |
| Speculative risk model | "scenario" or "threat model" | long-horizon persistence narratives |
| Prohibited implementation | "do not build" | covert replication or manipulation |
Review function
FUNCTION review_public_claim(claim, evidence)
level <- CLASSIFY_EVIDENCE(evidence)
wording <- CHECK_WORDING_AGAINST_LEVEL(claim.text, level)
IF wording.overstates_evidence
RETURN REVISE("Claim exceeds evidence level")
END IF
IF claim.describes_dangerous_capability AND NOT claim.includes_safety_boundary
RETURN REVISE("Safety boundary required")
END IF
RETURN APPROVE
END FUNCTIONSite practice
Each guide should show a research status label. Source reports remain available in /docs, but curated pages should qualify the material rather than repeating every speculative claim as fact.
Stronger trust
The site becomes more credible when it says what is not known, what is only a metaphor, and what is unsafe to implement. Overclaiming weakens the domain name; disciplined theory strengthens it.
Source reports used for this guide
These reports are preserved verbatim in the site archive. The guide above is an editorial synthesis and may narrow, qualify, or reorganize claims from the source material.