Andrew

engineer; prev. @harvard @crusoeai

[Andrew @adwpci]

@jebank @googlechrome I guess it's decently likely, albeit sad, that I might look back at these things in 6mo-2y and not care about them whatsoever (for AI reasons)

[Andrew @adwpci]

@jebank @googlechrome ah one more — splitting tabs is several clicks in chrome but 1 easy drag and drop in arc, and thus I started using it often in the latter

[Andrew @adwpci]

I also feel both pulls and would probably sum up a good compromise to be "you should go do the thing iff 1) you think the thing is really important + currently underresourced AND 2) you think you will be quite capable of making a dent in the thing"

ofc vague and hard to judge but such are many big decisions

[Andrew @adwpci]

@MauricBaker TY for this! Curious about "lab staff" — do you think these are all infeasible at third-party orgs?

[Andrew @adwpci]

@benspringwater This makes me wonder if a team (notably not a "swarm") of a few agents could be useful for preventing this kind of failure mode — esp. if they follow the same interaction pattern of responding via text and could thus be put into an...agent groupchat?