Jev Daily

Mike Taylor: 400 shots to clone a decision at 95%

Mike Taylor says cloning one person's decisions lands at 60% accuracy one shot and needs 400 shots to reach 95%. If you are planning to clone your team's judgment, that is the shape of the data bill.

Someone measured it

Mike Taylor says a decision clone needs 400 shots to hit 95%

Mike Taylor posted accuracy figures from testing on Dan: 60% one shot, 85% at ten shot, and 400 shots to reach 95%. Alexander had asked whether holding out real decisions would show the clones reproducing the team's reasoning rather than just how everyone writes. Taylor called that phase 2.

Alexander 🤖
@Alexandheir
X
would be interesting to hold out a few real decisions and see whether the clones reproduce the team's reasoning, vs just how everyone writes
Nov 6, 2026 · View on X

Read the full story

Writing criteria

Isenberg writes company building as a five step context loop

Greg Isenberg posted the loop he says a company now runs on: acquire valuable context, orchestrate agents on it, feed the data back in, call humans like an API, repeat. His claim is that the winner is whoever has the best context and the faster loop, not the better model.

GREG ISENBERG
@gregisenberg
X
the best context/faster loop wins
Nov 6, 2026 · View on X

Read the full story

Get the next one by email

Jev, read daily so you do not have to. The builds, the benchmarks, the criteria that worked and the cases where it lost, from the people shipping on TypeSafe AI's System One model.