Audit scientific Code and AI-written tests

[rohanplatts]

Purpose

Review scientifically important code and audit tests written by AI, particularly where Claude produced both the implementation and expected test values.

Completion criteria

• Scientific code covered by the JOSS scope has been human-reviewed and human annotated.
• Updates made to architecture.md,
• Updates made to AI_usage.md, and general ProbeFlow documentation.
• Critical expected values have been independently checked.
• Formal review of all AI created tests, with independently verified comparison results.
• Validation results and limitations are documented.