Asking AI agents to prove their uniqueness yields nothing but marketing word soup

altryne · x · 2026-09-14

altryne ran a fun experiment: he asked each of his agents (Instinct, Muse, Bot, Hermes) to explain why they specifically are the best at helping him. The result: marketing word soup — almost none could articulate a unique selling point, except Instinct and Hermes' 'switch the brain' feature.

Codex running on Astra produced so much boilerplate it needed a separate post. Takeaway: agent frameworks' self-descriptions are badly disconnected from their actual capabilities.

Related event: Agents Fail to Justify Their Own Uniqueness in Experiment(2 posts)→

Original post →

More from coding & agent

coding & agent channel →