README.md

OSL and Neuralese

README.md

Download original file

Original Markdown with line links

OSL-Core 0.5 Semantic Independence v0.1.0

OSL-Core 0.5 extends the Core 0.4 receipt-hardened runtime with separately implemented semantic verifiers for the finite numerical, contract, session, metadata-facing, and audit-intrinsic fragments. These verifiers operate on supplied source, receipts, and final state without importing or invoking OSLRuntime. Same-runtime replay remains supplementary and is not counted as independent semantic evidence.

This is a separate executable development release. It does not rewrite the frozen Operator Symbolic AI Language v0.5 / OSL-Core 0.3 package or relabel Core 0.4 evidence.

What changed

Verified release facts

Gate Result
Packaged source corpus 66/66
Accepted / expected rejection 42 / 24
Native Core 0.5 receipts 1,825
Independent semantic aggregate 66/66
Full test suite 168/168
Focused Core 0.5 suite 75/75
Discovery Bridge 3/3 programs, 71 receipts

The focused suite contains positive reconstruction, malformed-input, fully-rehashed semantic-mutation, rollback, integration, and language-hardening cases. It is a fixed finite gate, not exhaustive adversarial testing.

Independent coverage

The aggregate in reference/osl_semantic_kernel.py combines five bounded components:

Component Independently reconstructed surface
Numeric Spaces, vectors, operators, application, transport, projector and intertwiner checks
Contracts Five templates, feedback deadlines, explicit checks, finalization, per-domain rollback
Sessions Finite graphs, reports, lifecycle, history, terminal transitions, per-domain rollback
Metadata Secret commitments, metadata records, assertion counts, show bindings, per-domain rollback
Intrinsics The receipt-prefix report emitted by verify_receipts

The receipt-hardened protected-state kernel inherited from Core 0.4 remains responsible for source/lifecycle, packets, proofs, memories, capabilities, actions, and protected rollback checks. Exact same-runtime replay is retained as a separate compatibility and determinism signal.

Important limits

Core 0.5 does not yet independently establish:

The implementation and numerical profile remain within the Python and NumPy trusted computing base. Evidence is finite-corpus executable assurance, not a mechanized proof. SHA-256 binds bytes but authenticates no identity. OSL capabilities are symbolic and are not an operating-system sandbox. Trusted neural origin and relevance attestations remain trust roots rather than proofs of semantic truth.

Read CORE05_SEMANTIC_INDEPENDENCE_SPEC.md, CORE05_NUMERIC_PROFILE.md, CORE05_COMPATIBILITY.md, and CLAIM_BOUNDARY.md before making release claims. The native trace bundle is CORE05_TRACE_BUNDLE.jsonl; generated summaries are CORE05_CONFORMANCE_REPORT.json, CORE05_SEMANTIC_REPORT.json, CORE05_FOCUSED_REPORT.json, and CORE05_DISCOVERY_BRIDGE_REPORT.json.

Offline verification

Install the dependency ranges from requirements.txt, then run the packaged Core 0.5 entry point with ./verify_core05.sh on Bash or .\verify_core05.ps1 on PowerShell. Successful verification requires no network after dependencies are installed.

Other PDFs for this work

Original file and source record

This copy matches the recorded source file. Its publication here does not change the work’s review status.

SHA-256: cfcd1fa7039f445fdd63f8cdadf4b4316f96a4c20a29c4075ca076d0add1b1d3

Original source on GitHub

Back to OSL and Neuralese