About Veritas
Most AI tools are confident even when they're wrong. Veritas refuses to be.
The promise
Every question you submit runs through two independent AI verification passes from different model families — one from Google's Gemini line, one from OpenAI's GPT line. If they agree, you get the answer. If they disagree, a third tiebreaker pass with elevated reasoning settles it. If even that can't reach a 2-of-3 majority, Veritas marks the row as unverified and moves on.
This is the trade we made: fewer answers, but the ones we give are right.
What it's good for
- Multiple-choice study guides — paste 4 or 40 questions, get a clean letter answer key.
- Fact-checking a claim before you cite it.
- Quick reference questions where being wrong is worse than being silent.
What it's not
Veritas is not a chatbot. It doesn't make small talk, it doesn't write essays, and it doesn't have opinions. It's a verification engine wearing a website.
Honest limits
Verified does not mean infallible. Two AI models can share a blind spot. Treat Veritas as a strong second opinion, not as a substitute for a textbook, a teacher, or your own judgment — especially for anything that matters.