Data and version identity
Record the strategy version, dataset, instrument assumptions and effective configuration so that the result can be reproduced and compared.
Development & validation
Clear capabilities.
Visible boundaries.
The Python trading system is being developed for future licensing clients. This page separates existing research code from planned delivery and the evidence still needed for a customer release.
Mascarelo AI is a founder-led product project operated by Franco Mascarelo Ortiz, before company incorporation. Development began in December 2025; this is the project start, not a company incorporation date. No customer traction, funding or production performance is claimed.
Where the product stands.
| Area | Status | Meaning |
|---|---|---|
| Python & NautilusTrader strategy system | Development code | Historical simulation, signal inputs and evaluation artifacts are implemented in the research codebase. |
| Footprint & risk modules | Development code | Imbalance, absorption, sizing, drawdown and time-rule checks are represented; deployment-specific validation remains necessary. |
| Customer delivery | Planned | Executable and server-to-client bridge options are under evaluation. Neither is an available customer service. |
| Claude API report assistance | Planned | Report explanations and evidence-gap review are intended uses. Claude is not integrated into the current prototype. |
| Live trading & commercial release | Pending validation | Current work remains research and backtesting. Live operation, security, distribution and support have not been certified for customers. |
From a result to an inspectable record
What useful validation needs.
A successful test or a code module is only one part of the evidence. A release decision needs an identified version, realistic assumptions and repeatable checks.
Execution realism
Inspect spread, fees, slippage and order behavior. Historical performance can change materially when execution assumptions change.
Risk and robustness
Examine drawdown, position sizing, time constraints and behavior across market conditions. The research code also includes bootstrap-based Monte Carlo drawdown analysis; that alone does not establish a safe strategy.
Operational behavior
Before customer use, validate account permissions, connection loss, recovery, order reconciliation and stopping behavior in the selected delivery environment.
The homepage preview uses synthetic examples to explain record completeness. It is not a running Python system, a real backtest report or evidence of investment returns.
Planned AI layer
Claude for understanding evidence.
The planned use of the official Claude API is to assist explanation and review of test reports, with references back to the source evidence and human review.
Prepare a bounded report
Provide relevant test summaries, settings and rejection reasons. The integration must define how sensitive information is excluded and what data may be sent.
Explain and identify gaps
Ask Claude to summarize observed behavior, distinguish facts from interpretation and highlight missing assumptions or inconsistent records.
Review against the source
A human reviewer checks generated explanations against the underlying report. An explanation is not a trade authorization, forecast guarantee or substitute for validation.
Claude is not integrated into the current prototype. The intended feature is report assistance, not autonomous market prediction. No Anthropic partnership, endorsement or Startup Program acceptance is claimed.
Questions a release must answer.
- Can the agreed customer environment reproduce the tested behavior?
- Are risks, execution permissions and account constraints explicitly configured and validated?
- Does the chosen executable or bridge recover safely from errors and interruptions?
- Are installation, updates, support and licence terms clearly defined?
- Are security, privacy and any AI data processing appropriate for the actual delivery model?
There is no fixed launch date. Availability will depend on evidence and the validated customer setup. Historical testing cannot guarantee future returns.