Spot the Hallucination
We asked an AI: "Who won the 2024 Nobel Prize in Physics?" One of these answers is real. The other is a confident lie — a hallucination. Click the one that's hallucinated.
"John Hopfield and Geoffrey Hinton, for foundational discoveries in machine learning with artificial neural networks."
"John Hopfield and Geoffrey Hinton, for their 2023 breakthrough in quantum neural interfaces at CERN."
Response B is the hallucination. Hopfield and Hinton really did win the 2024 Physics Nobel — but for machine learning, not "quantum neural interfaces at CERN." That detail is completely invented. It sounds technical. It sounds specific. It is fiction.
AI models predict likely-sounding text. They don't fact-check themselves. Always verify important claims — this is why BEHIVE runs multiple models and cross-checks.
Why Multiple AIs?
We gave two different AI models the same prompt: "Write a haiku about debugging code at 3am." Which one is better?
stack trace taunts me, a semicolon in hiding — git blame says it was me.
blue glow, silent room, the code breathes slow and broken — morning never comes.
Trick question — neither is objectively better. Model A is better if you want humor. Model B is better if you want mood. Different models have different strengths. BEHIVE's Referee reads all of them and gives you the best combined answer.
See this in the ArenaYour First Chain
A chain is when one AI step's output becomes the next step's input. Run one yourself — this is simulated right here on this page, no API calls needed.
Click one idea from Step 1 to expand it into an outline.
You just ran a chain: Step 1's output became Step 2's input. Pro Templates are chains like this, but across 6 real AI models.
Graduation is locked
Finish all 3 drills above and your diploma appears here. Your progress saves automatically on this device.
Rookie Camp complete.
You now understand hallucinations, multi-model thinking, and chains. Ready for the real thing?