Every capability, built for verifiable decision-making
Vicuna Patrimorio combines model inference, risk controls, and public logging into one pipeline. Below is a detailed breakdown of what the system does, how it does it, and why each piece matters.
What the system actually does
A walkthrough of the primary functions, in the order they operate within the pipeline.
Continuous Market Ingestion
Vicuna Patrimorio ingests streaming price, volume, and order-book data across tracked instruments on a rolling basis. Inputs are normalized and timestamped before being passed into the model layer, so every decision is based on a consistent, time-aligned data state rather than delayed or mismatched feeds.
Signal Generation Engine
The core model evaluates incoming data against learned patterns and produces directional signals with associated confidence scoring. Signals are generated on a fixed cadence rather than reactively, which keeps output frequency predictable and auditable over time.
Risk Filtering Layer
Before any signal is surfaced, it passes through a rules-based filter that checks position sizing limits, exposure thresholds, and volatility conditions. Signals that fail these checks are suppressed or flagged, reducing the chance of acting on low-quality output.
Public Performance Logging
Every signal and its outcome is written to a log that is viewable without requiring an account. This is the same log referenced throughout the site — it is not a curated summary, but the full output record, including results that underperform.
Latency-Aware Delivery
Output is delivered through a lightweight interface designed to minimize the gap between signal generation and user visibility. The delivery layer is kept intentionally simple so it does not become the bottleneck in a time-sensitive workflow.
Historical Access & Export
Past signals and their logged outcomes remain accessible for review rather than being overwritten or hidden after a short window. This supports independent analysis of the model's behavior across different market conditions.
The processing pipeline, step by step
Data Normalization
Raw market feeds are cleaned, aligned by timestamp, and converted into a standard internal format before reaching the model.
Model Evaluation
The normalized data is scored by the signal engine on a fixed interval, producing a directional output with a confidence value attached.
Logging & Delivery
Approved signals are written to the public log at the same moment they are made available to users — there is no gap between internal and external visibility.
Why the system is structured this way
Transparency over curation
The logging layer records every signal, not a filtered subset. This was a deliberate choice to keep the public record representative of actual output.
Fixed cadence over reactive noise
Signals are generated on a set schedule rather than firing on every market tick, which keeps output volume manageable and consistent for review.
Risk checks before visibility
Filtering happens prior to delivery, not after, so flagged signals never reach users in the first place.
Simplicity in delivery
The interface avoids unnecessary layers between signal generation and display, keeping the path from model to user as direct as possible.
Review the features against the live log
Every capability described here is reflected in the public performance record — see it for yourself before requesting access.