Build Your Personal AI Benchmark
Join a cohort

The Lab: Build Your Personal AI Benchmark

Nufar Gaspar. Frontier Lab with The AI Daily Brief and Superintelligent. October 2026.

A new AI model or tool arrives every few weeks, with test scores and a week of hot takes. None of it tells you whether it is better at your work. A personal AI benchmark does. It is a small set of your own real requests. You give them to several AI options side by side, with the names hidden, and see which answer you would use.

This lab is self-paced, and you can do all of it without the recording. You leave with three things:

  1. A saved task list: four to six real requests you reuse every time something new comes out.
  2. A blind comparison: every option's answer to every request, names hidden, with your pick and one line on why.
  3. A decision: your AI lineup, or a verdict on something new.

One rule before you start: nothing private. You will paste these requests into many tools over the years. Use realistic made-up material with invented names and numbers. Prompt 1 writes it for you.

Parts 1 and 2 are the core, complete on their own. Then come an extra for early finishers and two optional parts for going deeper, with their prompts included.