[timfromhcs]

Summary of Changes

1. 200+ Qubit Scalable Simulation & Full-Register Sampling

• Path C (MPS-Native Sequential Sampling): Designed and implemented an exact batched sequential conditional sampler using right-to-left environment contraction and left conditional projections. Eliminates dense ^n$ statevectors, allowing exact sampling of 200–300+ qubit low-entanglement states (e.g. 300Q GHZ sampled with 1,000 shots in 0.27s with zero dense statevector allocation). Dense statevector conversion remains protected behind \MAX_STATEVECTOR_QUBITS = 20.
• Path A (Stabilizer Tableau Acceleration): Vectorized symplectic phase accumulation in _row_sum\ using an algebraic mod 4 formula (~10x speedup). Fixed a latent \self.MAX_RESAMPLE_SHOTS\ AttributeError on wide circuits. Enabled drawing full requested shots (e.g. 1,000 shots on 300Q GHZ circuits) with verified mathematical support.
• Path D (Selective Observables Request API): Added \observables\ request parameter to \execute()\ and CLI to evaluate explicit subsets of one- and two-qubit Paulis without (n^2)$ pair explosion. Automatically suppresses pair expansion for > 30$ unless explicitly requested.

2. Robustness & Mutation Harness Hardening

• Mutation Testing: Removed all shortcut
  eturn True\ fallback branches from \scripts/run_mutation_tests.py\ (mutants 1, 2, 5) and properly injected wide-counts fabrication in mutant 6. All 7 mutants killed (100.0%). Added meta-tests in \ ests/test_mutation_harness_meta.py\ ensuring fail-closed non-zero exit on broken baselines, surviving mutants, and harness errors.
• TV Bound Stability: Protected \hoeffding_total_variation_bound\ against integer-to-float \OverflowError\ on massive registers ( > 60$ bits).

3. Formalized OpenAI Math Corpus Audit (Phase 2)

• Pinned upstream snapshot at commit \1581d4cb. Parsed and audited all 372 manuscript families and 719 preprints.
• Generated full machine-readable manifest (\MANUSCRIPT_MANIFEST.jsonl, \MANUSCRIPT_INDEX.csv) and research reports in
  esearch/v2_1/openai_math/.
• Identified complexity barriers (Boolean oracle query separations $\neq$ classical circuit simulation).

4. Verification & Release Gates

• pytest: 220 passed, 65 subtests passed.
• CTest: 13/13 passed in Release configuration.
• Mutation Harness: 7/7 killed (100.0%).
• 200Q Acceptance Suite: 7/7 passed.
• Silent Truncation Audit: PASS.
• SPIR-V Shaders: 8/8 validated.
• Version Consistency: Checked and verified for 2.1.0 across all public surfaces.