It's an interesting idea, but in application it is not really usable.
Some prompts will get a working result one time, and an error message a second later. An example "current unix timestamp" will sometimes get me the current UNIX timestamp, and about half the time "Sorry, as an AI language model, I cannot generate current unix timestamp since it is dependent on the current time and date in your timezone. However, if you provide me with a specific time or date relative to today and a timezone to convert to, I can provide you with the corresponding time and date in the requested timezone."
If you hit enter more than once, you end up with the output of those two queries being mixed into each other. Using the same prompt as above "current unix timestamp" with enter hit twice with a second between, I get "I do not have access to real-timeSorry, data as and an cannot AI provide language current model timestamps,. I However do not, have if access you to give the me current a unix specific timestamp time as or it requires date real relative to-time interaction today with the and system a. timezone However to, convert I to can, convert I unix can provide timestamps to you readable with dates the and corresponding times time if and you date provide me in with the one requested. timezone." Which is two instances of the error message woven together awkwardly.
