i spent 2 maybe 3 hours setting up a process with the openai operator preview. after using the process twice this morning i was told by the operator that it could no longer do the process. i would never have gone done that work if i had known in advance that it would be wasted. there is no warning that such work would be wasted.
my question: why is openai not providing the warning?
Hey @kloro2006, totally get why that feels frustrating. Spending hours setting something up and then having it stop working isn’t ideal.
What’s happening is Operator is still in a research preview. So workflows can work once or twice, then hit a limitation and hand control back. It’s still evolving, which is why behavior isn’t always consistent yet.
On the warning part, you’re right. There isn’t a clear upfront signal today, mainly because the system can’t always predict when a flow will break. That’s a known gap.
For now, best approach is to treat these flows as experimental and test them a few times before relying on them.
If you want more context, you can check “Introducing Operator” for details on how it’s positioned today.
-Mark G.