kernel-reconstructed-evt-endpoint-exclusion-lower-leg
- Kind
- kernel-term
- Status
- checked
Supports: p(-1) > p(-3) for p = x^3-6x, admitted by crate::Kernel::add_declaration as Check.evt_endpoint_exclusion_lower_leg : Rat.lt Rat.zero (Rat.polyEval q 4 (Rat.ofInt -1)), where q = p - p(-3) (coefficients [9,-6,0,1]) built from the same [0,-6,0,1] integer coefficients the CAS's own extremum::polynomial_extremum(x^3-6x, -3, 2) certificate returns, from a Nat.le 1 14 witness -- 14 being the EXACT value the fully reduced polynomial computes to
cargo test -p axeyum-lean-kernel --lib rat_prelude::cas_evt_bridge_tests::tests::evt_endpoint_exclusion_kernel_checked -- --exact 2>/dev/null | grep -cE '^test rat_prelude::cas_evt_bridge_tests::tests::evt_endpoint_exclusion_kernel_checked \.\.\. ok$' Evidence notes
MUTATION-VERIFIED by this lane before registering: changing ONLY the kernel-side bound for this leg from the exact 14 to a wrong-but-plausible 16 (Nat.le 1 16 is itself a true proposition, just not the one the reduced term inhabits) made Kernel::add_declaration reject with TypeMismatch { expected: ExprId(1579401), got: ExprId(1579415) } and the test FAIL; reverted, it passes again. So the kernel term asserts what the CAS-derived arithmetic actually computed (p(-1)-p(-3) = 14), not merely something well-typed. The mutation and revert were both made and confirmed inside this lane's own worktree; git status was confirmed clean afterward.