job chat: if the question isn't clear, seek clarfiication

[josephjclark]

Sometimes users will accidentally post a question before they've finished typing. Or maybe a team member will run a test.

And the model will treat this as a legitimate question, even though there isn't anything for it to say.

For example, yesterday someone wrote "latency". Were they testing the model's latency? Did they type into the wrong window? I don't know. But the model generated a 200 word response. That took time, energy and money to generate. But there's no possible useful answer the model could have given, apart from "I'm sorry, I don't understand, can you rephrase the question pal?"