Modular, with explicit contracts
Each component has a defined role, inputs and outputs. Approved baselines are versioned so behaviour does not change silently.
Technology
Engineering for verifiability.
AI software has to be explainable, testable and safe when something goes wrong. These principles shape how WealthGenixAI builds its applications, starting with its flagship, WealthGenix Trading.
We publish principles and a simplified structure. Internal component designs, logic and security controls are intentionally not published.
How applications are built
From shared foundations to governed release.
Every application follows the same path. Each step has to be passed before the next one begins, and nothing reaches users without a separate approval.
- 01
Shared foundations
Every application starts from the same engineering, security and governance practices rather than from scratch.
- 02
Modular development
Components with defined roles, inputs and outputs, built and versioned so behaviour never changes silently.
- 03
Testing & verification
Automated tests, strict type checking and security analysis run before any change is accepted.
- 04
Governed release
Review gates and a separate approval before anything reaches users. Features that touch money, personal data or safety also need legal and regulatory review.
Architecture at a glance
Layered, with gates between the layers.
A simplified public view, from the interface down to execution. Gates must be passed before control moves to the next layer.
- L1InterfacePresents analysis and status. Designed not to authorise sensitive actions by itself.
- GateService boundaryAuthentication, authorisation and validation of every request.Being completed
- L2Orchestration & governanceCoordinates components under explicit rules and records what happened.
- L3Specialised analysisMarket context, strategy evaluation, opposing views, explanation.
- GateRisk & compliance checksMust pass before any simulated action proceeds.
- L4Simulation & paper ledgerWhere actions are rehearsed without real money.Enabled
- L5Live executionRequires legal, regulatory and security review and a separate decision.Not enabled
Engineering principles
How we build.
Each principle is chosen for the same reason: so that behaviour can be checked, explained and stopped.
Server-side authority
What is allowed is designed to be decided behind a service boundary, not trusted to the browser. This separation is being completed in stages.
Fail-closed defaults
If a check cannot complete or configuration is missing, the system stops rather than proceeding on assumptions.
Verification before change
Automated tests, strict type checking and static security analysis run before changes are accepted.
Untrusted-input isolation
External data is treated as untrusted and validated before any component relies on it.
Regulatory awareness by design
Jurisdiction-aware compliance checks are being developed as part of the core architecture, not added at the end.
Bounded automation
Automation inside clear limits.
Automation is useful only when its boundaries are explicit. Components act within defined permissions, and the most consequential capability — live execution — is not enabled at all.
Bounded
Defined permissions
Each component may only do what its contract allows.
Observable
Recorded behaviour
Significant actions are designed to leave a reviewable record.
Stoppable
Emergency stop
Activity can be halted immediately.
Resilient
Safe failure
Missing configuration or failed checks lead to refusal, not improvisation.
What this page does and does not claim
Parts of this architecture are implemented and verified; others are being built in stages. We describe them as design principles, not finished capabilities. WealthGenix Trading currently operates in simulation and paper-trading modes only.