Nufar Gaspar
A practitioner in the AI applications domain, focused on AI tool evaluation and workflow optimization. Co-host of an operator-perspective AI workshop series.
Timeline (last 90 days)
Nufar broke down the process of building a personal AI evaluation benchmark into five steps, with the first step—choosing the tasks to test—being the most decisive.
When selecting candidates, it is recommended to include no more than four, and at least one must be a baseline representing the current way of working.
In Nufar's tests, she and the AI judge reached agreement on only one task—price negotiation—and disagreed on all other tasks.
Decision outcomes can be summarized into three types: Switch, Split (mixed use across different scenarios), and Stay.
All Facts (4)
Nufar broke down the process of building a personal AI evaluation benchmark into five steps, with the first step—choosing the tasks to test—being the most decisive.
50%UnverifiedWhen selecting candidates, it is recommended to include no more than four, and at least one must be a baseline representing the current way of working.
50%UnverifiedIn Nufar's tests, she and the AI judge reached agreement on only one task—price negotiation—and disagreed on all other tasks.
50%UnverifiedDecision outcomes can be summarized into three types: Switch, Split (mixed use across different scenarios), and Stay.
50%