Paste a transcript. Watch it work.
Felarity scans the text for numeric, date, percentage, and named-quantity claims, attributes each to its speaker, and surfaces conflicting pairs. The output is a hash-chained Merkle JSON you can verify yourself on our trust page. Nothing leaves your browser.
INPUT Paste transcript, or pick a sample
0 chars
Format hint:
SPEAKER: text on each line. Sentences without a speaker prefix are attributed to "Unknown".
Runs entirely client-side.
RESULT Findings + signed chain
Idle
⊕ 0 words
👤 0 speakers
⚠ 0 contradictions
⛓ 0/8 chain
⌛ —
Paste text on the left, then click ▸ Find contradictions.
Sample run takes about 0.4s.
What this does: extracts dollar amounts, dates, percentages, and quantities; groups them by topical proximity; flags pairs where the same speaker (or the meeting record) commits to incompatible values. Real Felarity sessions add semantic NLI, cross-session comparison against the speaker's prior testimony, external source grounding (EDGAR/FDA/PACER), and acoustic analysis — none of which can run in 200 lines of browser JavaScript.