Discussion about this post

User's avatar
Kevin Kasaei's avatar

Agree that the traits did not change. What changed is where they are observable, and the take-home is the piece I would revisit. A 2027 planning data exercise was a strong judgment test when producing the artefact cost a candidate an evening. It now costs twenty minutes, and a lot of what you are grading is tool access.

The version that still works for me is to hand over the output rather than the input. Give the candidate a finished analysis with three defensible mistakes buried in it, tell them explicitly to use whatever tools they like, and score what they change and what they leave alone. Same trait, business acumen under pressure, but the model cannot do the part you are actually measuring, because the part you are measuring is knowing which answer is wrong.

One honest note on the first screening question. It is a good question and it is now among the most coached questions in existence. Every interview prep tool tells candidates to name an internal blind spot. I ask for the specific decision that blind spot caused them to get wrong and what it cost, because the coaching stops at the general answer.

No posts

Ready for more?