I've spent the afternoon testing ChatGPT's new consumer automation product - Agent. Where does it work, and where does it fall short? And how does it compare to Operator (and newer products like Perplexity's Comet)? My review 👇

It also often "overdid" simple tasks - but couldn't complete more complex ones. Ex. Agent took a full minute to find the date / time of a recent 1:1, and couldn't schedule a new one (below is 4x speed ⬇️) Comet found the same info and sent a new invite in <10 seconds.

ChatGPT's privacy and safety limits also get in the way. Agent agreed to make a birthday poster for @illscience, pulling the date from my calendar and a photo of him from the Internet. But the end result (took 13 minutes) redacted his name and picture for "privacy reasons"🤦‍♀️

What is Agent really good at? Getting a head start on non-time sensitive research or computation tasks. Ex. Agent created a DCF model of NVIDIA (in downloadable Excel format!) in 25 min. The formatting isn't exactly how I would have done it, but this is a big step forward.