Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Google product manager interview question - Write some code with an LLM tool that leverages a LLM to determine if the new owner of a domain is doing (a) same dodgy thing as prior owner that got flagged (b) different dodgy thing as prior owner but should be flagged (c) something completely innocuous (d) needs further review.


Please don't give Google ideas for more ways they can have an algorithm arbitrarily screw you over with no recourse, they're listening.


Well, current approach guarantees you’re getting screwed over. Any improvement is beneficial unless it blocks a better approach?


You're looking at this from the perspective of a haunted domain owner. And from that perspective your idea is fine.

A good technique to evaluate ideas though is to try and view it from different perspectives.

In this case from the owner of a non-haunted domain. Can you see any potential problem with your idea when viewed from that perspective?

Now, if there are potential problems, consider the relative sizes of the two groups. Do the benefits to one outweigh harm to the other?

This technique can be used every day with pretty much any idea.


The parents rules seemed to indicate only reevaluating the status of a haunted domain. I see nothing about evaluating a normal domain.


(Therefore, this has a one-way function of improving the status of haunted domains and why I think anything is better than nothing unless it blocks a better strategy.)


Follow up interview question. Update the code using your LLM code gen tool of choice that, when someone submits a complaint via an online form, feeds that complaint text back into your LLM to score it again. Points deduction if the candidate ever mentions informing the complainant of anything.


Why would they care?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: