# Reaction networks: effective equations and separate finite evidence

Selected reading distinguishes the older initial-state-dependent all-time bounds from the newer classwise eventual absorber. The formal scope covers only the older claim. The new construction orders tolerance, finite slopes, separation and class scale, with compactness-based choices; no practical numeric absorber or entry-time bound was extracted. Full source proof and kernel acceptance remain open.

[Detailed review](../../../../prototypes/reaction-network-2026-10-09-v1/report.md) adds 713 passing controls, effective-graph/rate-convention checks, exact supplied conservation and invariant-rectangle evidence, and conventional NetworkX/SymPy/SciPy comparisons. SBML/Catalyst import is unimplemented and physical fidelity is not established. The qualitative theorem does not replace finite numerical evidence.

[Opportunity 013 v2](../../../../opportunities/013-reaction-network-assurance/2026-10-09-v2.md) is a low-confidence specialist integration. Facility planning remains first. Source revision `fd4aeeb2ee4fc729c18d98444fed42fd0529eeeb`; four PDF pages inspected, precise selected extents recorded.