01
🎙
Capture Speech
A microphone near the podium streams live audio to Deepgram's medical-grade speech recognition (nova-2-medical), producing a live transcript with speaker identification.
02
🔍
Extract Claims
Claude AI continuously scans the transcript and identifies specific, verifiable medical claims — statistics, mechanism statements, efficacy claims, and regulatory assertions.
03
⚡
10 Parallel Searches
Every claim triggers simultaneous queries across 10 medical databases — PubMed, Cochrane, ClinicalTrials.gov, FDA, EvidenceMD, Semantic Scholar, Perplexity, and more.
04
🧠
AI Synthesis
Claude Sonnet weighs all returned evidence — prioritizing peer-reviewed sources — and renders a verdict with a plain-English explanation and source citations.
05
📺
Live Display
Verdicts appear on a projector screen visible to the audience in real time: green (accurate), yellow (contested), or red (inaccurate), with a full explanation.
06
✏️
Moderator Override
The moderator can override any verdict, add notes, manually fact-check a claim, and see a per-speaker accuracy score throughout the session.