Pedulli SRD Engine - Benchmarks
The Pedulli SRD Engine is Pedulli's universal SRD intelligence engine for free-form reasoning, domain reasoning, verifier-gated outputs, and AI acceleration. Every benchmark below reports a measured capability from the engine, exposed through measured demos and exact verifier-gated checks. A capability is listed as delivered when it ships with a measured benchmark and a published proof object (proof.records: SHA-256 + timestamp). Directions still under active exploration are labelled research. Engine internals stay sealed.
How to read this page
This is a record of what the engine delivers and the proof object that backs each measured capability. Delivered capabilities are exposed through measured demos, exact verifier-gated checks, and product APIs; research directions are stated as research.
The output satisfies the stated exact relation. The engine reasons to the result and an exact verifier confirms it, byte-for-byte.
When the exact relation is not met, the engine returns a precise diagnostic describing why - an honest, auditable signal, never a fake pass.
A direction under active exploration. Listed for transparency and labelled research; it moves to the delivered table once it ships with a measured benchmark and a proof object.
Per-domain verifier benchmarks
Each row is a domain reasoning surface running on the universal SRD intelligence engine. The check column states the exact verifier-gated relation behind the capability. "Delivered" rows ship with a measured benchmark and a published proof object; research rows are stated as research.
| Domain | Exact verifier-gated check | Returns | Status |
|---|---|---|---|
| verify.compression Compression | Exact .pdli restore with a SHA-256 roundtrip check against the original bytes. | accept / diagnostic | Delivered Measured + proof object |
| verify.sat SAT | SAT reasoning, exact verifier-gated: an assignment is checked against the clauses and returns accept or a clear diagnostic. | accept / diagnostic | Delivered Measured + proof object |
| verify.proof ATP / proof | Proof reasoning with each step exact verifier-gated, certifying validity end to end. | accept / diagnostic | Delivered Measured + proof object |
| verify.rsa RSA relation | Exact verifier-gated bigint p*q == N check. Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only. | accept / diagnostic | Delivered Measured + proof object |
| verify.aes AES relation | Exact verifier-gated AES-GCM re-encrypt check. Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only. | accept / diagnostic | Delivered Measured + proof object |
| verify.reverse Reverse engineering | Exact verifier-gated trace replay: a candidate model is confirmed to reproduce the observed traces. | accept / diagnostic | Delivered Measured + proof object |
| verify.mdp MDP / game | Exact verifier-gated value and policy equivalence under a stated symmetry. | accept / diagnostic | Delivered Measured + proof object |
| verify.trading Trading (diagnostic) | A diagnostic harness: out-of-sample window, cost, slippage, drawdown and leakage checks. No ROI and no Sharpe - strictly diagnostic. | diagnostic | Delivered Diagnostic surface |
Every domain follows the same contract: the engine reasons deterministically and an exact verifier returns accept or a clear diagnostic. Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only. AGI and post-LLM intelligence are Pedulli's research direction, labelled research, not delivered results.
proof.records - measured records
Each delivered benchmark is backed by a proof asset: the measured artifact, its SHA-256 digest, and a timestamp anchored on the Bitcoin chain via OpenTimestamps. The timestamp proves existence and measurement priority. It does not disclose engine internals - proof without disclosure.
The exact input, the engine's output, and the verifier-gated result (accept or diagnostic) captured as a fixed artifact.
A SHA-256 hash of the artifact so anyone can confirm the published record is the one that was measured, byte-for-byte.
An OpenTimestamps proof on the Bitcoin chain fixing when the digest existed. Existence and priority only - internals stay sealed.
proof.records entry (shape):
{
"domain": "verify.<domain>",
"result": "accept" | "diagnostic",
"artifact": "<measured artifact>",
"sha256": "<64-hex digest of the artifact>",
"timestamp": "<OpenTimestamps .ots, Bitcoin-anchored>",
"disclosure": "sealed - proof without disclosure"
}
Research directions - labelled research
These are directions under active exploration on the universal SRD intelligence engine. They are listed for transparency and labelled research. They join the delivered table above once they ship with a measured benchmark and a published proof asset. AGI and post-LLM intelligence are Pedulli's research direction, stated as research.
Extending free-form and domain reasoning to additional encodings and surfaces. Labelled research until it ships with a proof object.
Scaling typed, verifier-gated reasoning jobs to higher capacity. Listed for transparency, labelled research.
AGI and post-LLM intelligence are Pedulli's research direction. Stated as research, not as delivered results.
How capability becomes delivered. A capability moves from research to delivered only when it ships with a measured benchmark and a published proof object. Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only. AGI and post-LLM intelligence remain Pedulli's research direction, labelled research, not delivered facts.
Submit a bounded job and watch the universal SRD intelligence engine reason deterministically and return a verifier-gated accept or a clear diagnostic. Every delivered capability is backed by a SHA-256 + timestamp proof object.
Verifier-gated - accept or diagnostic - Bitcoin-anchored proof - internals sealed