SRD Engine - the universal SRD intelligence engine

Submit a witness.
Get a verifier-gated verdict.

Pedulli SRD Engine is Pedulli's universal SRD intelligence engine for free-form reasoning, domain reasoning, verifier-gated outputs and AI acceleration. In this trial you pick a domain, supply a witness, and the engine returns a verifier-gated verdict: accept or a measured diagnostic. Public capabilities are exposed through measured demos, proof objects and verifier-gated product APIs.

pick domain -> supply witness -> verifier-gated check -> accept | diagnostic

YOU SUPPLY - SRD VERIFIES - ACCEPT OR DIAGNOSTIC

Every capability shown here is measured and verifier-gated; anything not yet measured is labelled research. Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only. SRD internals stay sealed - proof without disclosure.

Live trial - witness submission

Pick a domain. Paste or upload a witness.

The form posts your witness to the verifier-gated product API at /api/srd/verify. The endpoint is a business surface in front of the engine; when it is gated, the engine returns a measured diagnostic explaining what it would verify, never an unverified pass.

Step 1 Choose a verification domain
Step 2 Supply your witness

Paste the exact bytes/JSON of your witness, or upload a file. The engine verifies verify.compression: the supplied .pdli must restore to the original and the SHA-256 must match.

Or drop / choose a witness file
.pdli container, clause file, proof transcript, ciphertext, key material, trace log
You supply the witness - the engine returns a verifier-gated verdict. Cybersecurity domains are authorized-scope only.
Step 3 Verify the witness
// ready - pick a domain and supply a witness

The engine returns a verifier-gated verdict: accept or a measured diagnostic. SRD internals are never exposed in public replies.

Back to the SRD Engine overview ->

What the trial actually does

A universal engine with verifier-gated outputs

For every domain the rule is the same: you bring the witness, the universal SRD intelligence engine verifies it against an exact relation, and you get a verifier-gated verdict - accept or a measured diagnostic. Every public capability is measured; anything not yet measured is labelled research.

01

You supply the witness

Factors for an RSA modulus, a key for an AES-GCM ciphertext, an assignment for a clause set, a restored .pdli, or a proof transcript. You bring the candidate.

02

The engine verifies it exactly

A real p*q == N bigint check, a re-encryption that must reproduce the exact AES-GCM ciphertext, a SHA-256 roundtrip, a supplied-clause SAT check, or each proof step in order.

03

Verifier-gated verdict

If the residual is zero it returns accept. Otherwise it returns a measured diagnostic that says what did not pass. Outputs are verifier-gated, so an unverified run is reported as research, never as a pass.

Delivered & measured
  • The universal SRD intelligence engine, with verifier-gated outputs you can check yourself
  • Deterministic and local: no tokens, no context window, no LLM
  • Measured: a capability ships only with a measured benchmark and proof object
  • Returns accept or a measured diagnostic - never an unverified pass
Scope & measurement discipline
  • Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only
  • AGI and post-LLM intelligence are the active research direction, labelled research - not delivered
  • verify.mdp is a correctness check under a stated symmetry, reported as measured equivalence
  • verify.trading is a diagnostic only - it reports measured behaviour, no ROI or performance claims

Domain reference

What each domain verifies

Each domain is a separate exact relation, verifier-gated and measured. Click a row in the trial to load its hint, then supply the matching witness.

verify.compression - the supplied .pdli must restore to the original and the SHA-256 must match
verify.sat - your supplied assignment is evaluated against your clauses
verify.proof - each supplied proof step is verified in order
verify.rsa - supplied factors are verified, p*q == N (verifier-gated, authorized-scope only)
verify.aes - your supplied key must re-encrypt to the exact AES-GCM ciphertext (verifier-gated, authorized-scope only)
verify.reverse - your candidate model must replay the observed traces
verify.mdp - decision / game: value & policy equivalence under a stated symmetry (measured correctness equivalence)
verify.trading - diagnostic only: OOS, cost, slippage, drawdown, leakage (measured behaviour, no ROI claims)
proof.records - SHA-256 + timestamp proof assets

FAQ

Quick answers

How does the engine handle RSA / AES / ECC?
Cybersecurity verification and research surfaces are verifier-gated and authorized-scope only. For verify.rsa you supply the factors and the engine verifies p*q == N. For verify.aes you supply the key and it re-encrypts to confirm it reproduces the exact AES-GCM ciphertext. The engine returns a verifier-gated verdict on the witness you provide, within authorized scope.
What do I get back?
A verifier-gated verdict: accept or a measured diagnostic. If the witness satisfies the exact relation, you get accept. Otherwise you get a diagnostic describing what did not pass. Outputs are verifier-gated, so an unverified run is reported as research, never as a pass.
What is verify.mdp?
verify.mdp is a measured correctness check: decision / game value and policy equivalence under a stated symmetry. It confirms two formulations agree under that symmetry and reports the result as measured equivalence.
Does verify.trading make ROI claims?
No. verify.trading is a diagnostic only: it reports measured out-of-sample behaviour, costs, slippage, drawdown and leakage on what you supply. It makes no ROI, Sharpe or performance promises.
The backend says it is gated - is that broken?
No. The verify endpoint is a verifier-gated product API, a business surface in front of the engine. When it is gated, you receive a measured diagnostic explaining exactly what the engine would verify for your domain, never an unverified pass. Capabilities ship only with a measured benchmark and published proof object.
Where does AGI fit in?
AGI and post-LLM intelligence are Pedulli's active research direction, labelled research - not delivered. The universal SRD intelligence engine is the delivered product: free-form reasoning, domain reasoning, verifier-gated outputs and AI acceleration, exposed through measured demos, proof objects and verifier-gated product APIs.
Do you expose the engine internals?
No. SRD is self-sealed: proof without disclosure. You can verify the result yourself, but the internals stay closed and there are no absolute-security claims.

Bring a witness. See a verifier-gated verdict.

The universal SRD intelligence engine verifies what you supply against an exact relation. Pick a domain above and submit a witness, or read the engine overview first.

Submit a witness -> SRD Engine overview

Universal SRD intelligence engine - verifier-gated outputs - measured capabilities - authorized-scope cybersecurity - self-sealed