01
Abstraction ability
AI excels at logic, yet fails certain logic tests. Why?
Read the articleFind candidates
Evaluate your candidates on AI's weaknesses: creativity, recognition of new patterns, understanding of ambiguity, commercial negotiation, and more.
Marie Claude
Application for the Head of Growth position
Test results
Candidate
Claude 3.5
GPT-4o
Gemini 1.5 Pro
Claude 3.5GPT-4oGemini 1.5 Pro#1
out of 54 candidates
Creativity
0
100
25
50
75
18
32
41
87
#2
out of 54 candidates
Understanding ambiguity
0
100
25
50
75
24
37
45
82
#1
out of 54 candidates
Sales negotiation
0
100
25
50
75
21
35
42
78
We have scientifically identified AI's weaknesses
02
Adapting to uncertainty
LLMs handle a stable world well. When the market shifts, they stay frozen.
Read the article03
Emotional understanding
Reading an emotion in context is essential at work — and AI gets it wrong.
Read the article04
Commercial negotiation
LLMs recite Fisher & Ury, but lose money on every deal.
Read the articleTest · Abstraction · Easy#ARC-6150A2BD
A simple, fast workflow
For each campaign, pick your tests:
For each role, an AI agent recommends the most relevant test or tests.
Share the provided link
A public QR code to take the tests, or email invitations — your call.
Receive a comparative report
We compare your candidates to each other, and to the different AIs. You then make an informed choice.
AI has blind spots that science reveals more clearly every day.
Our team spends most of its time reading and producing scientific research on AI's limits. The robustness of our tests comes from thousands of research papers — mostly Chinese, American and European — that we manually analyse and whose protocols we replay.