AI should draft the strategy—not hide the rules

Thor Oracle is designed as a research copilot. It can translate a plain-language thesis into a deterministic rule draft, but the resulting strategy remains visible and editable.

Generation and execution are different jobs

AI may suggest an indicator, threshold, or allocation. Once approved, deterministic code should evaluate those rules consistently. A language model should not improvise paper orders from an opaque conversation.

Every draft needs a challenge

Thor Oracle identifies the implied market regime, likely failure mode, concentration risk, and the evidence still missing. The user sees both the strategy and the argument against it.