Yutori's Maximillian Piras argues agents need measurement, OpenAI shares In Harmony and Cognition shows Devin testing itself with GPT-6 Astra. Together they point to evaluation and accountability for autonomous systems.
Mousepower: agents that can’t be measured, can’t be managed.
Maximillian Piras of Yutori presented Mousepower at AI Engineer. He frames measurement as a prerequisite for managing agents in production, a talk now available on video.
In Harmony
OpenAI published In Harmony on YouTube. No additional description was provided with the listing.
Cognition helps Devin test its own work with GPT‑6 Astra
Cognition is using GPT-6 Astra inside Devin to improve testing.
The goal is to help Devin verify its own work and demonstrate that it works. OpenAI says the approach should help engineers review less code and ship more.
