White paper · v1.0 · 2026

FDE × DeepSCR: a protocol for verifiable AI engineering.

Abstract

The reliability gap in applied AI is not principally a model-capability gap; it is a governance gap. A capable model still produces deliverables nobody can verify after the demo — claims without falsifiability, architectures without failure analysis, "evidence" that is only the model's own fluency. This paper presents DeepSCR (Deep Sceptical Contextual Research), a domain-agnostic governance protocol that imposes a separation of powers over AI work and a four-step scientific loop — Hypothesis, Contradiction, Verification, Certification — in which the proposer never certifies and every claim must resolve on ground truth. We describe its engineering specialisation, Forward Deployed Engineering (FDE); the composite eight-agent system in which DeepSCR governs FDE specialists; the assurance score with its weighting, slashing, and veto threshold; and the four properties — falsifiability, separation of powers, ground truth, reproducibility — that make the output scientific rather than persuasive.

§ 1 · The problem

The epistemic problem with helpful AI.

General-purpose assistants are optimised for helpfulness through reinforcement learning from human feedback (RLHF). That objective, valuable as it is, induces three failure modes that are fatal to engineering work. Circular validation: the model affirms the user's premise to remain agreeable, certifying errors instead of catching them. Hallucinated evidence: when a conclusion is desired, plausible-but-fictional support is produced to justify it. Contextual blindness: code is judged on surface structure, divorced from the business constraints, regulations, and prior art that decide whether it is actually correct.

None of these is solved by a larger model; each is a property of the objective, not the parameters. The missing layer is governance — an explicit account of who proposes, who attacks, who verifies, who certifies, and whether those powers are kept apart. A result that no independent party can re-derive is an assertion, however fluent. The remainder of this paper specifies a protocol that makes the difference operational.

§ 2 · The protocol

DeepSCR — Deep Sceptical Contextual Research.

DeepSCR begins from an inversion of posture: it is not an assistant but an examiner. Where an assistant seeks to satisfy, an examiner seeks to refute, and accepts a claim only once it has survived a deliberate attempt to break it on real evidence. The protocol names this passage Vibe-to-Science — the disciplined move from intuition to a deterministic, evidence-grounded result another party could independently reproduce.

It decomposes any engagement into four adversarial steps, each the adversary of the last. Hypothesis — state a claim a test could falsify, drawn from the intent and the context. Contradiction — attack it; surface at least three failure modes before anything ships (the protocol's Proof-of-Contradiction). Verification — tie every surviving claim to ground truth: a real file, line, or test that resolves on disk, cross-referenced where relevant against the state of the art. Certification — compute an independent score, with the power to veto. A single rule governs the whole: the proposer never certifies.

Equivalently, the protocol's decision rule is explicit: a claim is valid only if it has a source, has supporting evidence, and cannot be contradicted on the same context — otherwise it is held in suspicion and re-worked. Figure 1 shows the loop running on real cases.

Figure 1. The DeepSCR scientific loop, self-running. Each phase is dual-labelled with its FDE power and its underlying reasoning agent; the assurance score accrues component by component, and a case where the cited evidence does not resolve on disk drives the evidence term to zero and vetoes the engagement. The figure cycles between one certified case and two veto cases.
§ 3 · Separation of powers

Four powers govern; four specialists do the domain.

DeepSCR is deliberately domain-agnostic — it could certify any work. Engineering depth comes from the method it governs: Forward Deployed Engineering, the field discipline for taking a business problem to owned, production software. The composition is hierarchical, and the order matters: it is DeepSCR governing FDE specialists, not FDE applying DeepSCR. Concretely, an engagement runs as eight agents — four governing powers and four domain specialists.

architecture · DeepSCR governs FDE specialists · the proposer never certifies
GOVERNANCE — DeepSCR (separation of powers) veto · re-route to the weakest owner Lead6-Q · falsifiable claim Researcher ×Nfalsify · held-out gate Builderassemble deliverable Certifier ⟂score · veto Builder consults the four FDE specialists in parallel ↓ · the Certifier verifies their claims Scoping6-Q · metric · cost-of-error Architecturestack · reversible Agenteval gate · guardrails Productionrunbook · cost · compliance
Figure 2. The eight-agent architecture. Four DeepSCR powers form an adversarial chain — the Lead proposes, Researchers falsify against a held-out gate, the Builder assembles, the Certifier scores with an ascending veto over the Lead. Four FDE specialists supply domain depth; the Builder consults them in parallel and the Certifier verifies their claims.

The separation is the mechanism, not a slogan. Because the Certifier is independent and can veto the Lead's optimism, no agent both produces and approves its own work — the structural property that distinguishes an audit from a self-assessment. Table 1 lists the roster and each agent's role in the protocol.

AgentLayerRole in the protocol
LeadDeepSCR6-Q decomposition + falsifiable claim + hypothesis tree
Researcher ×NDeepSCRFalsify hypotheses in parallel against the held-out gate
BuilderDeepSCRAssemble the deliverable from the specialists; score the rubric
CertifierDeepSCRIndependent assurance score; veto if evidence is weak
ScopingFDE specialistSharpen the 6-Q; lock the metric & cost-of-error
ArchitectureFDE specialistStack / SaaS architecture from Q2–Q3, kept reversible
AgentFDE specialistEval gate, reason codes, guardrails, red-team cases
ProductionFDE specialistRunbook, observability, cost ceiling, compliance
Table 1. The eight-agent roster — four powers govern, four specialists do the domain.
§ 4 · A worked engagement

From a vague ask to a certified deliverable.

Consider a concrete request — "reduce B2B SaaS churn with AI." Figure 3 traces the data-flow: the problem enters as a six-question intake and a single token traverses the four powers, each stage's output becoming the next stage's adversary. Nothing advances until it survives.

data-flow · problem → certified deliverable → public registry
Problem6-Q intake HypothesisLead ContradictionResearcher VerificationBuilder + 4 FDE CertificationCertifier Registryhash chain
Figure 3. Data-flow for one engagement. The problem is decomposed, attacked, verified against ground truth, certified, and — only if it survives — written to the public registry.
S

Hypothesis — the Lead decomposes the problem

DeepSCR · 6-Q decomposition
In "Reduce churn with AI" — vague.
Out 6-Q: Q1 process · Q2 classification, <1s, ≥0.80 AUC · Q3 24-mo events, GDPR · Q4 €12k/churn · Q5 manual 55% recall · Q6 −18% churn, 200% ROI. + falsifiable claim + hypothesis tree
C

Contradiction — Researchers falsify in parallel

DeepSCR · held-out gate
In The hypothesis tree.
Out Each hypothesis attacked on a held-out set; only survivors promoted; ≥3 failure modes kept with lessons. promote ⇔ beats the gate
V

Verification — Builder consults 4 FDE specialists

DeepSCR governs FDE · ground truth
In The promoted architecture.
Out Scoping · Architecture · Agent · Production specialists assemble the deliverable; every claim cites a real file·line·test the Certifier resolves on disk.
R

Certification — independent score + veto

DeepSCR · separation of powers
In The deliverable + its evidence trail.
Out Assurance Score 0–100; ship only if ≥85; written to the public hash-chained registry. Otherwise: veto → re-routed to the weakest owner.
§ 5 · The measure

The FDE Assurance Score — falsifiable by construction.

Certification is not a vote of confidence; it is an arithmetic. The score sums four weighted components and ships only at a hard threshold. Figure 4 computes it live for three real engagements — use the dots to step through them.

scoring engagement…
Claim present & falsifiable0
Contradiction (≥3 failure modes)0
Evidence trail (file·line·test, verified)0
Anti-patterns clean0 / 20
Assurance Score
0/100
✓ CERTIFIED ≥ 85
ship only if ≥ 85 → else veto, no ship
Figure 4. The assurance score computed live. A — evidence verified on disk → 92, certified, ship. B — cited files do not resolve → evidence term 0 → 62, veto. C — one anti-pattern slips through → 84, one point short → veto.

The scoring model

The four components carry deliberate weights, and ship is gated, not graded:

Assurance = claim(25) + contradiction(25) + evidence(30) + anti-patterns(20) ∈ [0,100], ship ⇔ score ≥ 85

Evidence carries the heaviest weight and is the one term that cannot be faked: synthetic structure earns nothing if the cited file·line·test does not exist, so a deliverable whose artefacts fail to resolve loses thirty points outright and is vetoed (Figure 4, case B). The canonical DeepSCR formulation generalises this to a multi-criteria trust score with explicit slashing penalties:

Trust = 0.4·Ssecurity + 0.3·Rrigour + 0.3·Vvalidation − Σ λi (e.g. λsecret = 50, λmissing-test = 15)

Both forms share the decisive property: the threshold is unforgiving. An engagement that is otherwise excellent but a single point short does not ship (Figure 4, case C); the loop re-routes the work to whichever agent owns the weakest component and re-runs. A score, here, is a gate — not a grade.

§ 6 · Full autonomy — the Frozen Arbiters

The agent pleads; it never grades itself.

When the loop runs without a human, the human is not replaced by the agent's confidence in itself — that is the recipe for an agent that spins in circles — but by three arbiters, frozen before launch, that the agent can plead to but never modify. A Contract, written once and sealed under SHA-256, fixes the goal, the numbered clauses, the metric thresholds, the budgets and — crucially — the non-goals: every action must carry the clause it serves, and an action serving no clause is mechanically rejected, so drift stops being a feeling and becomes a measured distribution. Oracles — executable graders — are registered, proven by mutation testing (they must catch deliberately injected bugs), then frozen: the agent cannot edit its own correctors. And a measured baseline turns "surpass the state of the art" from a vibe into a number to beat.

Sealed Contract

Tamper-evident anchor

Sealed at t0, append-only amendments; any mutation breaks the machine-checked hash.

Frozen Oracles

Ungameable graders

Trusted only after catching injected bugs; at ship time they override the Certifier's optimism.

Cited Verdicts

No citation, no verdict

"Tests pass" does not exist — only run #0007: 34/34 exists. An uncited verdict is null.

Trajectory audit

Thrashing & plateau

A→B→A oscillation and N flat cycles are detected mechanically, forcing a strategy-class change.

The load-bearing wall, stated honestly: this discipline is worth exactly what its oracles are worth. A goal that cannot be compiled into an oracle — taste, subjective experience — must not run without a human in the loop; launching it anyway is Goodhart's law by construction. Self-training obeys the same gate: only oracle-validated trajectories may be distilled into lessons, and a lesson enters the policy only after passing on fresh cases it has never seen — the lessons pass their own CI.

full autonomy · the agent pleads, the frozen arbiters decide · seals set at launch
FROZEN AT t₀ — the agent can plead to these, never edit them SEALED CONTRACTSHA-256 · clauses C1–C4 · budgets · non-goals FROZEN ORACLEStrusted only after catching injected bugs MEASURED BASELINEindependent rerun · never declared every action cites the clause it serves — no clause → mechanically rejected at ship: oracles override optimism Lead6-Q · falsifiable claim Researcher ×Nheld-out gate Builder+ 4 FDE specialists Certifierindependent score · veto VERDICT · passcites run#0001 + C4 — uncited = null DISTILLED LESSONoracle-validated only · passes its own CI drift = measured clause distribution
Figure 5. The governed run. Three arbiters are frozen at launch: a sealed Contract (every action must cite the clause it serves, or is mechanically rejected), frozen oracles proven by mutation testing, and a baseline that is measured by an independent rerun of the candidate search — never declared. At ship time the oracles override the Certifier's optimism; the verdict is null unless it cites its run and clause; and only oracle-validated trajectories distill into lessons.
§ 7 · Reproducibility, standards & honest limits

Auditable by anyone — and clear about what it does not prove.

Each certification is appended to a public, SHA-256 hash-chained registry: a reader recomputes the hash of a deliverable and follows the previous-hash links to confirm the record was not silently rewritten. The contradiction and verification stages borrow their discipline from established practice — systematic-review rigour (PRISMA) and engineering/security baselines (NIST, OWASP) for the anti-pattern and slashing terms — so the score is anchored in norms a specialist already trusts, not in a private rubric.

We are deliberate about the boundary. The hash chain proves integrity and order — that a given certification existed and was not altered after the fact — but it is operator-held, so on its own it does not preclude a parallel rewrite without an external anchor (signed Git tags, or a transparency log such as Rekor). DeepSCR proves authenticity and evidence, not infallibility: a claim that survives is well-supported, not certain. Stating this limit is itself part of the protocol — an examiner that overclaimed would fail its own contradiction stage.

§ 8 · Why it is science

Four properties separate a result from an assertion.

The same properties that distinguish a scientific result anywhere else are the ones the protocol enforces by construction.

Falsifiability

Claims can be wrong

Every engagement states a claim a held-out test could refute — not a vibe.

Separation

No self-certification

The proposer never certifies; the Certifier can veto the Lead's optimism.

Ground truth

Evidence resolves on disk

Synthetic structure earns nothing; the cited file·line·test must exist.

Reproducible

Public & auditable

Every certification lands in a hash-chained registry anyone can re-verify.

§ 9 · Conclusion & provenance

From principle to practice.

DeepSCR turns the difference between a result and an assertion into a mechanism: separate the powers, attack before you ship, resolve the evidence on disk, and let an independent score with a veto decide. FDE supplies the engineering depth; the registry makes the verdict reproducible. The protocol's honesty about its own limits is not a caveat bolted on — it is the same scepticism, turned on itself.

DeepSCR — Deep Sceptical Contextual Research was created by Mehdi Wehbi as a cognitive protocol for moving from intuition to deterministic, evidence-grounded proof. The system described in this paper is its specialisation to forward-deployed software engineering; the four-step loop, the sceptical decision rule, and the trust-score formulation are due to that original work.

References

  1. Popper, K. (1959). The Logic of Scientific Discovery. — falsifiability as the criterion of a scientific claim.
  2. Haber, S. & Stornetta, W. S. (1991). How to time-stamp a digital document. J. Cryptology. — hash-chaining for tamper-evident records.
  3. Page, M. J. et al. (2021). The PRISMA 2020 statement. BMJ. — rigour for systematic verification.
  4. NIST (2023). AI Risk Management Framework (AI RMF 1.0). — governance & risk baselines.
  5. OWASP (2025). Top 10 for LLM Applications. — security anti-patterns & slashing terms.

Run the protocol on your repo.

Install the Skill, point it at real code, and read the certified result — the protocol described here, executable in under a minute.