Hey PH - maker here. The story that pushed me to build this: one Tuesday I spent two hours chasing a Bitrix24 API method that an AI assistant described in detail, with code examples and confident tone. The method doesn't exist. I'd been working from one model's confident answer. That night I ran the same question through three models. Three different method names came back: - ChatGPT: crm.item.userfield.add - DeepSeek: crm.userfield.add - Gemini: userfieldconfig.add Only Gemini had the right one for smart processes. The other two were either invented or borrowed from a different part of the API that doesn't apply. All three answered with the same confident tone. No hedging. That's the whole product. Concordance runs your question through three independent models simultaneously. A fourth model synthesizes them and outputs a consensus score (0–100). High = the models converged, well-covered territory in training data. Low = they diverged, someone's guessing, worth verifying. A small benchmark I ran (60 questions, half general / half narrow technical): general questions clustered at median 95% consensus. Narrow technical questions sat at median 33%. The signal lives in the bottom half - that's where one model is quietly wrong. It doesn't pick a winner. It can't - that fourth model doesn't know the right answer either; no LLM does. It just makes disagreement visible, so when models split, you know to verify before relying on a single answer. Built solo, for a problem I kept hitting myself. Free tier: 3 queries - try it on something you've been trusting one model for. Founding tier: $9/mo for the first 100 people, price locked for 3 years. Try it on something you've been trusting one model for - and tell me what you find. Longer write-up here: https://dev.to/concordance_ai/i-...

Concordance

Three AI models on the same question. See where they split.

Three AI models on the same question. See where they split.