01
Compare two AIs
Compare two labelled authored answers to the same email, then judge what matters.
2 minutesCompare answersBuilds — try AI on real office work.
Free, no sign-up. Each tool opens with something useful already on screen.
Runs on worked examples — how this works
02
Is this ready for an agent?
Get a preparation score and the boundaries to fix before a pilot.
4 minutesAssess a workflow03
What does AI actually cost?
See the cost per task, month and year for one repeatable workflow.
2 minutesCalculate the cost04
Spot the hidden instruction
Find the untrusted instruction and compare bounded and unbounded responses.
2 minutesInspect the items05
Let an AI read a bank account
Connect a sandbox bank and watch the agent answer through six read-only MCP tools.
3 minutesAsk the agentTesting long context? Try Lost in the Middle.