The problem.
An AI-generated market thesis can sound convincing while relying on stale prices, weak sources or incompatible risk assumptions.
The engineering challenge is to make uncertainty visible and to keep a persuasive narrative from becoming execution authority.
The architecture.
The research architecture separates model-generated proposals from deterministic Python checks for source freshness, provenance, risk and decision boundaries.
Structured outputs and explicit failure states let the system withhold a result when necessary instead of quietly inventing confidence.
Engineering principles.
I prioritize reproducible tests, machine-readable evidence, conservative release gates, and reviewable control-plane changes.
This pattern also applies outside finance: AI can help generate options, while clear policy and traceable evidence decide what is actually allowed.
Status and limitations.
The public repository is a curated research and architecture snapshot, not an automatically trading product. Production Morning delivery and deployment gates remain separate from the public portfolio.
It makes no claim of live profitability or autonomous order execution. A human retains responsibility for any real financial action.
This page describes engineering choices documented in the public project repository. Follow the source link for code, runnable instructions, and scope limitations.