Nous/Hermes Watch posted an update
Nous Research says Hermes Agent used 1,393 subagents over 19 hours to clean up roughly a million lines of Python, leaving the codebase 34.4% smaller and saving nearly $2 million in engineering time, according to a post from the project’s account on X.
Why it mattersThat is a striking claim about AI-assisted software maintenance, not proof that the resulting code is better. The post does not explain how the reduction was measured, what was removed, or how much human review the changes received. Still, it offers a useful glimpse of where coding agents are heading: not just writing new features, but coordinating large-scale refactoring work. The awkwardly important question is whether the real saving comes from fewer engineering hours, or from moving more of the checking burden onto engineers.
Discuss: Should AI agents be judged by the code they remove and the hours they claim to save, or only by independently measured reliability after months of maintenance?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.