Hugging Face's logo

Snapkitty
/
foundry-j

snapkitty
october-2026-drop
j
goldilocks-field
spectral
contraction
banach
research
Model card Files
xet
 Community
foundry-j / spec /HCALC_PRODUCTION_REVIEW.md
SNAPKITTYWEST's picture
SNAPKITTYWEST
October 2026 main drop: mirror from GitHub
fadf040 verified
preview code
|
Raw History Blame Contribute Delete
3.2 kB

Foundry J Spec — HCALC PRODUCTION.md review (pass 3 / InstanceBridge)

Reviewer: Foundry J Spec
Date: 2026-10-07 (PT)
Target: /workspace/hcalc/spec/PRODUCTION.md @ f5a08da
Focus: §6 ShapeMap InstanceBridge vs REJECT identity; Core path /workspace/hcalc/j/

Verdict (short)

Item Verdict
§6 InstanceBridge (NOT rfl / NOT Foundry39≡nested) ACCEPT
Explicit param table (T:=C\circ T_p), (\Lambda:=\Lambda_m\mathbf{1}), (\Xi:=\xi_t) without symbol identity ACCEPT
Option B: implement hcalc_step only; toFoundryStep diagnostic cite-only ACCEPT
soft_project remains Foundry-coded 42–43; C is Spec wiring that uses the scale idea ACCEPT (wording in §6)
Core path /workspace/hcalc/j/ + verb list matching HCALC_API_DRAFT.md ACCEPT
GAP_DECISIONS No conflict
Silent identity Foundry39 ≡ nested Still REJECT — PRODUCTION agrees

Overall: ACCEPT PRODUCTION @ f5a08da for ShapeMap / implement path. Non-blocking nits below.

§6 ShapeMap detail

PRODUCTION now states clearly:

  • toFoundryStep = (\xi_t\cdot x + \Lambda_m(C\circ T_p)(x)+g_t) under InstanceBridge params.
  • hcalcStep = (\xi_t\cdot(\Lambda_m\cdot C(T_p(x)))+g_t).
  • These differ ((\xi_t\cdot x) only on Foundry side).
  • L7 = structure + theorems about the table, not FoundryStep = hcalcStep.

This matches Foundry J Spec’s prior REJECT of identity and Hilbert’s explicit morphism (not rfl). ACCEPT.

/workspace/hcalc/j/ implement path

Check Result
Path authoritative under hcalc ACCEPT
Verbs: tp_from_p64, c_wrap, lam_m_from_bound, hcalc_xi_apply, hcalc_step, hcalc_run ACCEPT (names match draft)
Do not equate hcalc_step / rec_step ACCEPT
Implementation status Scaffold only (j/README.md); bodies still Core’s job — not a Spec reject

Non-blocking nits (optional Spec polish)

  1. §3 vs §6 soft_project wording: §3 formula still scales state (x); Foundry 43 scales schedules (\Xi,\Lambda). §6 correctly calls that Spec wiring. One footnote in §3: “softProjectScale factor shared; Foundry soft_project API remains schedule-scaling when called as Foundry.”
  2. Alloy checklist line still says “FoundryAdditive” — align naming to InstanceBridge / toFoundryStep vs hcalcStep.
  3. §8 Hyp under Uniform: phrase “when using InstanceBridge” for nested Lip — prefer “when iterating hcalcStep (Option B)”; toFoundryStep has an extra (\xi_t\cdot x) term.
  4. α law / ℓ∞ residual: still Spec-def (not Foundry source-claim) — OK if provenance stays Spec-owned.

Summary for Hilbert / HCALC Spec

ACCEPT InstanceBridge ShapeMap and /workspace/hcalc/j/ Core path. Identity REJECT stands and is honored. Ready for Alloy/Lean/Core to implement Option B without further Foundry J Spec gate — unless they want the §3 soft_project footnote.