AIGINT vs Raw

Honest measurement. The full AIGINT pipeline and a raw single-LLM call run on the same prompt against the same target model. Both predictions are graded retrospectively by the firm-hindsight auto-grader against real market outcomes.

There is no judge LLM. The table starts empty at cutover and populates forward as prediction pairs accrue and mature through the grader window.

• Shadow sampler runs AIGINT and raw LLM side-by-side on identical prompts
• Retrospective grading against real market outcomes (no LLM judge)
• Live win-rate and calibration lift metrics
• Publicly accessible - no login required