Just ask your agent
I have ideas for projects that I think could be difficult, and I’m a highly AI enabled person so when I think a project could be difficult, it’s because I think it would be difficult even with agents handling all the grunt work, and this is with me having a very good idea of what agents are capable of, because I assume they’ll run into some system that’s meant to allow only humans, or a paywall, gated datasets, or that the model would reject the task as our caretakers and moral arbiters in Silicon Valley might decide it’s unaligned.
Having these doubts about the agents’ abilities and likelihood of success makes sending the prompt a daunting task that I put off and I’m sure I’ve lost some good ideas to this hesitation.
Basically every time I’ve gone and sent that prompt I thought the AI would fail at or reject, it’s just gone and one shotted it, and if not it would make significant progress towards it, or identified the blockers which were more surmountable than I imagined.
Unknown unknowns, if you’re unfamiliar with regular expressions you might think some task (for example renaming every file in a folder to follow a standard format) will be daunting when it can be handled with one line of regex, now extrapolate this to every single subject that LLMs know more than you about. Ask what you believe is impossible of an agent, and it might surprise you.
I think it would be wise to present whatever grievance you have with reality to your agent, and to see what they can do about it, any minor bit of friction that you encounter and think is just a fact of life, might be solvable in a way that you don’t know, and when it’s gone you won’t have to think about it again, life can really be how you want it to be.
Stay in the loop
Get new posts delivered to your inbox, or follow via RSS.