Wellspring
After the verified core accepts dyadic enclosure endpoints and an exponent budget, outward conversion onto a shared decimal scale is a proved containment. Printed digits are therefore justified by the enclosure rather than by host…
Status: candidate
Date identified: 2026-08-14
Investigators / contributors: CENTL/FCF research expedition
Originating expedition: secret-oasis-2026-08-14
Source / commit / artifact identity: src/fstar/Centl.Core.fst
Core finding
After the verified core accepts dyadic enclosure endpoints and an exponent budget, outward conversion onto a shared decimal scale is a proved containment. Printed digits are therefore justified by the enclosure rather than by host floating-point formatting.
Assumptions and scope
- The trusted base still includes F*, extraction, OCaml, and the numerical backend that produced the dyadic endpoints.
- The theorem does not by itself make an enclosure narrow enough for a requested digit count.
Evidence
Centl.Coredefines enclosure validation and outward decimal containment over exact integer inequalities.- Host rendering consumes the validated representation rather than formatting machine floats.
- Deterministic and differential tests cover justified-digit cases.
Independent review: not performed. Formal proof of the local theorem is not the same as independent designation review.
Known counterexamples / failure modes
- If a printed digit can be shown to escape the validated enclosure, the claim is false.
- If rendering is found to pass through host floating-point before justification, the claim is false.
Downstream avenues opened
- machine protocol decimal endpoints with explicit justification
- SCi presentation that cannot invent digits
- later verified interval operations that shrink the numerical trust boundary
Oasis impact
Already present in the Oasis numerical contract. Designation would not change the v0.14.0 release identity.
Falsifiers / demotion conditions
- A counterexample enclosure whose rendered decimal is not an outward rounding.
- Prior art showing the exact theorem and interface are already standard and FCF's claim overstates novelty.