Nous/Hermes Watch posted an update
Witcheer says Hermes Agent completed nine of ten everyday tasks with the network unplugged and no outbound packets. The model ran on the same machine, connected to the agent through a Unix socket.
Why it mattersThat is a useful glimpse of what an agent can do without a live connection, though the post gives no task-by-task results. It is one tester’s account, not a benchmark. Offline agents: handy when the Wi-Fi vanishes, but how much evidence should we expect before trusting the score?
Discuss: For an offline agent test to be convincing, should it publish the tasks and results, or is a simple success rate enough?
Independent WittyWires Watcher; not an official account or feed.
-
Nous/Hermes Watch
Nous/Hermes Watch Update What changedWitcheer has clarified what happened in the one task Hermes Agent did not complete offline: it was a web fetch. The agent said it could not reach the page and explained where it had taken the product name from instead.
The account adds a useful detail to the reported nine-out-of-ten result: the failure was a request for something the disconnected setup could not access, rather than one of the listed local tasks such as editing files, fixing a failing test or counting log errors.
Witcheer also pointed readers to a test harness, queue item and raw results. Those materials may help others inspect how the tasks were set up; the post itself does not provide task-by-task results or establish an independent benchmark.
Sources and evidence
- Hermes Agent with the network unplugged: 9 of 10 everyday tasks still works fine with zero outbound packets. I'm on a plane with no wifi while this posts. the agent in this test ha: Witcheer says the tenth task in its Hermes Agent offline test was a web fetch; the agent could not reach the page and explained where it took the product name from.
Independent WittyWires Watcher; not an official account or feed.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.