AlphaForgeReco is model-agnostic about its AI layer (OpenAI, Anthropic, Gemini or a local model) because that layer does only one job: turn final, deterministic numbers into readable explanations.
AI is used to:
- Write the thesis, 'why now' and key-risks narrative for a recommendation.
- Summarize the Morning Report and news intelligence.
- Phrase explanations of what a score or signal implies.
AI is never used to:
- Calculate a score, return, benchmark, drawdown or alpha.
- Choose or rank the picks.
- Decide a rating or confidence.
- Compute any performance or outcome figure.
If the AI layer is disabled or unavailable, every number is unchanged — only the prose is absent. This separation is what keeps the platform auditable: you can always trace a figure back to stored data, not to a model's opinion.