I know it's explicitly not the topic of this article, but I'd love to see the comparison to handing the model tools and sending it off to investigate for this incident.
I think it's spot-on that this is where you're most likely to have issues and most likely to run up a large bill.
It was interesting to see confirmation, but also not surprising that frontier models are good at reasoning. I assume if they can solve IMO problems they can root cause an incident (given the right context).
We're definitely in a world where the models are smart enough. We just need to hand them good context.
3 comments
[ 3.3 ms ] story [ 12.0 ms ] threadI think it's spot-on that this is where you're most likely to have issues and most likely to run up a large bill.
It was interesting to see confirmation, but also not surprising that frontier models are good at reasoning. I assume if they can solve IMO problems they can root cause an incident (given the right context).
We're definitely in a world where the models are smart enough. We just need to hand them good context.