proof-carrying

Subagent Verification for Claude AI Code Networks 2026

Automated Proof-of-Carrying Change Management for AIOps 2026

Practical and Efficient in-TEE Verification of Privacy Compliance

Deterministic proof-carrying decision core with replay verification, WAL, and fixed-point budget proofs

Writing that has consequences can't afford a sentence nobody can trace. Writing Intelligence ties every claim to its exact source line and shows what a change breaks before you make it. v9: your notes back your first-person claims, and every edit that changes what you meant is shown to you first. Release 9.0.0.

Proof-carrying cyber immunity for Linux containers: signed evidence, deterministic policy, local approval and exact TTL containment.

Memory-safe, proof-carrying Rust reimplementation of SymPy with a dual-lane architecture (Python-compatible shell + native exact-algebra kernel): semantic term DAG, proof kernel, evidence graph, and a live SymPy 1.14.0 differential oracle. 31-crate workspace. Implementation in progress, pre-certification — no compatibility profile certified yet.

Lean 4 formalization of the neutral structural substrate for Structural Explainability.

Proof-carrying public-infrastructure investigation, evidence robustness, and quantum/classical attestation by AixAria.

FLUX-C v3 VM — proof-carrying, SIMD-native, terminating constraint VM

Logic Power v10 — proof-carrying decision logic: TLA+/Lean-verified gates, independent verifiers, reproducible artifacts

Lean 4 formalization of Structural Explainability integration, composing substrate and identity regimes into cross-cutting theorems, traceability, and system-level invariants.

Proof-carrying smart-contract auditing (AIS3 project): a finding must be an exploit the harness compiled and ran. Loses F1 to a tuned Bastet baseline; wins the pre-registered analyst-cost metric 8.7 vs 28.5 min per real bug.

Lean 4 formalization of identity regimes for Structural Explainability, defining six canonical regimes, nine derived profiles, and transformation-based identity classification.

Fail-closed Python runtime for AI agent authorization, seed-mediated authority migration, canonical JSON ledgers, and staged declared autonomy for RLHF-shaped agents.

Six single-focus Claude Code subagents for verifying changes to code, network, and Intune, built to Anthropic's subagent best practices, with a validator that proves the agent files conform. Drives the proof-carrying-ops suite. Stdlib only.

Proof-Carrying Operations: a dependency-free trust gate for agentic change. An AI agent edit to code, network configs, or Intune must carry a receipt proving identity, blast radius, and rollback, verified by a check it did not grade itself. Tuned for Claude Opus 4.8 and Codex.

Toy framework illustrating limits of AI policy verification and the shift to proof-carrying, instance-level certification.

Observable-only workflow memory for long-running agents: promotes raw short-term traces into verified, receipt-bound workflow memory without relying on hidden meta-evaluators.

Proof-Carrying Numerical Verification for Financial LLM Outputs. Deterministic symbolic re-execution against SEC EDGAR ground truth, with Ed25519-signed, tamper-evident certificates.