research: finite LE semantic evaluator with replayable C2 evidence - #145
research: finite LE semantic evaluator with replayable C2 evidence#145mountain wants to merge 3 commits into
Conversation
Source freeze completeThe evaluator implementation is now remotely frozen at Pre-reveal checks:
The preimplementation held-out commitment remains |
Held-out reveal complete
Local verification is green: 18 workstream tests, 2 default CI bridge tests, and the post-reveal manifest. Disposition: EXPAND within the exact frozen fragment, with no broader surreal/transseries claim. |
|
CI #1809 is green: Python 3.10 and 3.14 compatibility jobs passed; the Python 3.12 full-test-and-build job passed. The research gate is complete and the PR is being marked ready for review. |
Closes #144.
Outcome
EXPAND within the frozen finite LE fragment. This PR upgrades the carrier compiler's syntax-level C2 classification into exact, replayable semantics for a bounded real single-exponential rational-rate task family.
Implemented:
t = exp(-N/q) -> 0+;Commit–reveal
eedef97f8071ea3da687a805b7ea3e8385aa49c0829f2c962a3c6bc8ec1816555df657d450233e14fa6d77a1efb5be36d6bc02e6q=3, exact1/5, normal form1/5 - t/6 + O(t**2), cancellation jump 4, valid 69-step replayVerification
sympy.limitClaim boundary
The same-information SymPy baseline also computes every numerical answer. The software gain is explicit semantics and replayable C2 discharge, not a raw computability/speed advantage. Surreal numbers play no substantive runtime role here. No general LE/transseries/hyperseries/surreal, symbolic-height, arbitrary nesting, complexity, Core, Theory Map, or Public API claim is made.