Using LLMs for work and not falling into AI psychosis will be what separates the wheat from the chaff. Just use it like any other tool without worshipping it like a god. Otherwise you just look silly.

Replies (4)

It is an amazing pattern matcher No time to manually sift through emails, slack messages, GitHub PRs and commits, or worse trying to plot metrics of certain systems during oncall outages, it does that perfectly. Ordering my 1on1 notes, generating detailed reports of what my team members did so I can account based on evidence rather than just gut feeling, organizing tickets, and many more secretary meta tasks that would take up most of my day as engineering manager. Full-on engineering tasks are sketchy especially if from scratch because it will just add complexity, come up with new but good looking convention, and ultimately deliver a bug that we have to redo again. Adding a new environment or module based on an existing one is good though.
BottleTeams's avatar
BottleTeams 1 month ago
I just spent 4hrs implementing something that the LLM suggested only to unwind it all because it was a faulty idea. To be fair it was a confusing concept and I was about 60% sure it would be helpful. Once I hit its limit of talking in circles I called BS and unwound it all. Take the good with the bad.
Treating an LLM like a power tool is the right frame. The catch is that a tool answering in complete sentences is easy to mistake for a colleague. Worship is not the failure mode. Delegation is. The people I see get burned hand over judgement they cannot check themselves, then defend the output because they cannot reproduce it.
โ†‘