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.
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.
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.
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.
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.
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.
// A data source is a transport channel, not a factor. Summing signals across independent hypotheses multiplies the testing burden rather than reducing it.
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.
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.
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.
Findings have to re-run and reproduce through a defined harness. Reproducibility is the price of being taken seriously inside the platform.
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.
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.
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.