Human intelligence league
Take timed challenges that show how well you direct AI, catch its mistakes, and deliver work people can trust.
This person's decisions clearly improved on what the AI produced alone.
Choose your fight
Start with a quick 12-minute check, or go deep and build a complete record of your work. Every level asks for real work, not AI trivia.
Pilot result for learning and development; hiring and credential use are not active.
Can you spot when AI sounds right but is wrong?
A fast diagnostic of verification instinct. Review an AI-generated market brief, isolate unsupported claims, and decide what is safe to use.
Pilot result for learning and development; hiring and credential use are not active.
Turn an ambiguous brief into a defensible answer.
Work through a constrained client brief with a controlled AI assistant. We look at how you plan, hand work to the AI, improve drafts, and check facts — and whether you spend your limited AI budget where it matters most. Clever-sounding prompts alone won't score.
Pilot result for learning and development; hiring and credential use are not active.
Do the work — not a quiz about the work.
A realistic research assignment with deliberately unclear parts, sources that disagree with each other, and a set limit on how much AI you can use.
What makes the difference
HILArena looks beyond the final answer to how you frame, direct, verify, and own the work.
Decide what question you are answering, what your limits are, and what proof you need.
Delegate deliberately instead of asking the model to "do it all."
Check important claims, catch weak evidence, and match confidence to what the work supports.
Make the final call and explain the judgment you added beyond the AI.
What you take away
See the decisions that made the work more accurate, useful, or safe.
Spot the checks you skipped, the evidence you overtrusted, and where effort went to waste.
Leave with a focused skill to practise before you step up to the next challenge.
Pilot scores are for learning and development. Broader uses appear only on evaluation versions cleared for them. See the evidence levels.
Your next benchmark
Take the first challenge, inspect the evidence, then climb.
Start with 12 minutes
