The platform

An architecture designed to measure its own value.

Most systems assume their components help. ASQIOS is stacked so the contribution of every layer can be isolated and measured against a control — and so the research integrity infrastructure can never be quietly bypassed.

Portfolio hierarchy

Four portfolios. One question each.

Each portfolio adds exactly one layer to the one before it, so the difference between them is the measurable value of that layer — not a tangle of confounded effects. Tap any portfolio to expand.

PORT-A

Factor baseline

Pure factor model.

The control. A systematic factor model (STRAT-001) with no AI layer at all. Every other portfolio is measured against this baseline.

+ DETAIL
PORT-B

Factor + veto

Baseline plus an AI veto.

Adds a constrained supervisory veto on top of PORT-A. The question it answers: does an AI veto add value, or destroy it? The veto is engineered to fail closed, never open.

+ DETAIL
PORT-C

AI analyst

A scoring AI analyst (CAA-002).

A fuller analyst that scores names under strict rules: it never excludes on data anomalies, never re-runs the Shariah screen, and never reads routine insider activity as a signal. Exclusions require hard evidence of fraud, insolvency, or going-concern risk.

+ DETAIL
PORT-D

Event monitor

A news-event monitor (NED-002).

Tests which event type carries edge — earnings surprise, insider pattern, contract award — not whether a data source does. Each hypothesis must meet its own sample size and survive its own multiple-testing correction.

+ DETAIL

// A data source is a transport channel, not a factor. Summing signals across independent hypotheses multiplies the testing burden rather than reducing it.

The honesty spine

The cage is built before the capability.

Three pieces of infrastructure sit beneath all research. They are deliberately small, deliberately rigid, and non-negotiable — the part of the system that exists to stop the rest of the system from fooling itself.

IMMUTABLE STORE

Data that cannot be quietly edited

Inputs are captured and fixed, so a result can always be traced back to exactly the data that produced it. There is no rewriting history to make a finding look better.

TRIAL LEDGER

Every trial recorded, win or lose

Verdict criteria are logged before the test runs and the outcome is logged after. Post-hoc stopping rules are forbidden, so the ledger is evidence rather than a curated highlight reel.

REPRODUCTION HARNESS

A result that cannot be reproduced does not count

Findings have to re-run and reproduce through a defined harness. Reproducibility is the price of being taken seriously inside the platform.

FAIL-CLOSED DESIGN

Errors resolve to caution, not optimism

When a component cannot parse or decide, it returns an explicit error state — never a quiet default that happens to flatter the result. Ambiguity is never allowed to drift positive.

Current data environment

Research runs on open, public sources: regulatory filings, public end-of-day prices, public macro data, and Shariah-universe proxies. This is enough to produce valid rejections — and explicitly not enough to produce institutional acceptance.

◦ R-2A · OPEN DATA · REJECTION-CAPABLE
Data environment

Built to grow into stronger data — never to pretend it already has.

The platform is structured so that the move to point-in-time institutional data is a deliberate, gated decision rather than an assumption baked into today's results.

Until that gate is opened, the honest position is that ASQIOS can rule strategies out, refine its machinery, and prove its own reproducibility — while reserving any institutional-grade verdict for institutional-grade data.

The structure only works because the governance is real.