How Nyalai works
Submit a strategy.
Or commission a verdict.
Two paths. Same discipline. Whether you build models or you allocate against them, the process below is how a Nyalai verdict is produced.
Path A
You built the model.
For quantitative researchers, systematic managers, and model authors seeking an independent adversarial verdict on their own work.
- Submit the one-page thesis.Prose only. The economic claim, the class of edge (speed, behavioral, structural), the instrument universe, the intended regime, the sizing envelope, the expected failure modes. Public and non-editable after submission.
- Provide the resolved-prediction ledger.A documented history of predictions the model has made, with realized outcomes. Timestamped, reproducible, in a format we can audit end to end.
- Receive the Verdict-form.Binary output: CALIBRATED or NOT-CALIBRATED. Full audit trail, per-gate diagnostic, verdict scope statement. Published under CC0.
Path B
You are about to allocate.
For institutional allocators, family offices, endowments, and OCIOs commissioning an independent second opinion on a manager or a model before deployment.
- Describe the target and the scope.Who or what is being validated. Which strategy or model. What regime and instrument universe. Whether the target has consented to submit their thesis and ledger (Nyalai does not validate on stolen data).
- Nyalai coordinates the submission.We contact the manager or model author to collect the one-page thesis and the ledger. The commissioning allocator is named on the resulting Verdict-form unless a confidentiality arrangement applies.
- Receive the Verdict-form and the diagnostic.Binary verdict, full audit trail, and a plain-language executive summary suitable for an investment committee, a board, or a governance memo.
Requirements
What Nyalai needs to render a verdict.
A one-page thesis
Prose. What the strategy exploits, the claimed edge class, the target regime. This is a hard admission requirement per Doctrine Section 9.0. No thesis, no verdict.
A resolved-prediction ledger
Documented history of predictions with realized outcomes. Minimum size typically N ≥ 200 for MTL-2 consideration. More for higher tiers. Reproducible from source.
Instrument scope
Public equities, listed derivatives, liquid credit, spot and forward FX, listed commodities. Private markets are currently out of scope (see Doctrine Section 1.4).
Disclosed exploration history
The number of hyperparameter configurations, feature sets, and training-window variants tested during development. Undisclosed exploration is a Gate 3 failure by default.
Sovereignty separation
At MTL-3 and above, the submitting party and the validator are structurally distinct entities. Nyalai does not co-develop the strategies it validates.
No time pressure
Verdicts take as long as they take. Deadline-driven submissions violate a hard constraint (Doctrine Section 8, item 5) and will be declined.
Input integrity
Sequence commitment.
Nyalai never requests permission to publish a case, or to use it in any downstream decision, before a verdict is rendered. Consent for publication or downstream reference is collected after the verdict is rendered, on a document the submitter reviews. Refusal of downstream use does not alter the verdict itself.
This is not a mental-state promise. It is an order-of-operations commitment that Nyalai controls and any third party can audit. Its purpose is to prevent a class of validity failure in which a submitter, aware of what will happen with the result, revises the input to shape the outcome. A submission contaminated by anticipation of downstream use is a false submission, and a verdict rendered against it would carry error we introduced ourselves.
1. Submission is frozen first
Timestamp and hash are recorded before any conversation about downstream use begins. The audit trail is immutable from this moment.
2. Gates run without stake exposure
The strategy author is never told, during evaluation, what downstream consequences might attach to the verdict.
3. Consent comes after the verdict
Publication, private client disclosure, aggregate research, or external reference are all conditioned on consent collected on a document the submitter reviews after the verdict is fixed.
Per Refusal Doctrine Section 12.4. This is a discipline of validity, not virtue.
What you receive
The Verdict-form.
A signed, timestamped document structured as a medical report: findings first, methodology available on request, references cited for audit. Two layers: the upper layer states the verdict in institutional vocabulary and names the primary risk category the strategy exposes the allocator to (shortfall, drawdown, liquidity, or embarrassment). The lower layer supplies the technical evidence: DSR value, BSS and ECE numbers, regime-by-regime Sharpe, bootstrap confidence intervals, guided-search correction magnitudes. Triangulation across risk type, business context, and regulatory frame.
Every Verdict-form is published under Creative Commons Zero. Anyone can inspect, reproduce, or contest the reasoning from the same ledger. See two live examples: the first Nyalai refusal Synapse v0.3 (NOT-CALIBRATED) and the illustrative reference case Reference Predictor R-01 (CALIBRATED).
The full methodology, gate specifications, thresholds, and primary-source bibliography are documented in the Refusal Doctrine v0.3.3.1.
Timeline and cadence
How long a verdict takes.
Clean ledger
Two weeks
A one-year, reproducible ledger on a defined universe. Standard case.
Ledger reconstruction
Four to eight weeks
When the submission requires cleaning, completion, or re-derivation of the prediction history.
Revalidation cadence
Per MTL tier
MTL-3 quarterly, MTL-4 biannual, MTL-5 continuous. Every verdict is time-bounded.
Trust is not a result of validation.
It is the product of the rigor with which we accept to contradict ourselves.
Sebastien Assohou · Nyalai · Founder