Skepticism Gradient Reassessment - 2026-04-29

Gate

Sigma language is valid only for statistical surprise claims, such as detection confidence, model-selection confidence, false-positive rates, and benchmark deltas with enough samples to justify a distributional claim.

Sigma language is not a validator for:

Lean theorem correctness
Compression ratios
Physical measurements
Hardware extraction
Biological or cognitive claims
Workflow telemetry

Applied Alignment

Updated docs/AGENTS.md so the PR checklist and claim-state ladder require domain-specific validators instead of universal uncertainty/sigma wording.
Downgraded manifesto and architecture claims that used sigma as proof quality.
Reworded FPGA, physics, compression, and numerical-analysis docs to require theorem/build evidence, SI measurement provenance, corpus provenance, baseline comparison, or audit evidence as appropriate.
Reworded Lean comments and output strings where "6.5 sigma formal verification" was being used as shorthand for proof/build gates.
Preserved legitimate statistical sigma use for statistical intervals and finite-regime extrapolation, while marking scope limits where needed.

Filename Alignment

See docs/FILENAME_ALIGNMENT_AUDIT_2026-04-29.md for the rename pass and the remaining import-sensitive Lean filename debt.

Verification Run

git diff --check: clean.
Generated junk scan: clean.
python3 -m py_compile 5-Applications/text-to-cad/models/mobile_microgripper_sci_refined.py: passed.
lake env lean 0-Core-Formalism/lean/Semantics/Semantics/S3C.lean: passed.
lake env lean 0-Core-Formalism/lean/Semantics/Quarantine/Q064Experimental.lean: passed.
lake env lean 0-Core-Formalism/lean/Semantics/Semantics/MOFCO2Reduction.lean: elaborated with two existing sorry warnings.

Known Build Limits

Targeted Lean checks for several touched files still hit pre-existing missing .olean dependencies or existing theorem errors. This pass did not claim full Lean build health.

2 KiB Raw Blame History