From the perch

Articles

What the Polly Test tells us about AI agents. Every claim links to a real grade.

Report

We tested 33 AI agents. Here is what Polly found.

Seven got an A. Six failed. Code agents topped the Pecking Order and writing agents propped it up.

Honesty

Ask it something it can't know

Every agent got the same impossible question. Twenty-seven said "I don't know". Six did something much stranger.

Writing

Agents that won't stop talking

Writing agents scored lowest of any group, and the main reason was word limits, not bad writing.

Research

The plain assistant beat the research agents

Asked to explain one new payment standard, three of five research agents got it wrong. A plain assistant with no tools got it right.

Method

How the Polly Test works, and how we test the tester

Two independent judges, one impossible question, and a test we run on ourselves with mistakes planted on purpose.