A continuous, multi-source scanning layer that samples public market data across the tracked universe on a fixed cadence. Because no single exchange's feed is treated as automatically authoritative, readings are cross-checked for staleness and disagreement before anything downstream relies on them.

Rather than deeply analyzing every asset continuously, a scoring layer ranks the full universe using low-cost signals, then applies deeper technical analysis only to the top-ranked candidates -- a deliberate discovery-first design, not an execution decision on its own.

Deterministic, no-AI indicator computation -- moving averages, momentum, volatility, relative strength -- applied consistently and transparently. Every value is derived directly from real market data; nothing is inferred by a black-box model.

Entry conditions are explicit, versioned rules that must all agree before a signal is accepted -- trend alignment, momentum confirmation, volume confirmation, and data-quality checks. See the Strategies page for the specific families of conditions used.

An independent gate that re-verifies every safety constraint on every proposed trade -- notional limits, position-count limits, instrument eligibility -- regardless of how confident the upstream signal is. This layer never trusts the strategy layer; it re-checks everything itself.

Order sizing respects each instrument's own exchange-defined precision, submissions are idempotent (safe to retry without duplicating an order), and every filled entry is immediately paired with a protective exit order before anything else happens.

A supervising process verifies the trading system is alive and reconciles local state against the exchange's own authoritative records on every cycle and after any restart -- because exchange-side events (like a protective order triggering) can happen independently of the system's own process.