21 October 2026
Experiment tests whether AI-built software can replace a $43,000-a-year tool
First reported
Latent Space ran this on .
- 72 builders submitted attempts to replace a speaker-management tool that costs $43,000 a year.
- The winner chosen by human judges ranked 19th on the AI judge's leaderboard, so the two rankings disagreed sharply.
- Critics argue that swapping out established enterprise software and maintaining custom replacements is expensive.
Our read
Only Latent Space among the 30 tracked newsletters covered this experiment, which we first reported on 2026-10-21. Our archive's earlier story on this thread is from 2026-08-27, about testing whether AI explanations of its own behavior actually work.
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
How it was covered
Latent Spaceswyx & Alessio
The 72 submissions, judge disagreement, build-versus-buy tradeoffs, and maintenance costs