HackerTrans
TopNewTrendsCommentsPastAskShowJobs

hhmc

no profile record

comments

hhmc
·6 bulan yang lalu·discuss
git worktrees are the canonical solution
hhmc
·6 bulan yang lalu·discuss
What is a ‘wicked’ problem?
hhmc
·8 bulan yang lalu·discuss
> For 99% of users, we gladly accept these more "invasive" anti-cheats because it means less cheating in the games we enjoy.

The modal user likely doesn’t even know anti-cheat exists, and if they did, wouldn’t care at all. They just want to play the game.
hhmc
·9 bulan yang lalu·discuss
> Most of us were able to earn money to buy a pizza or some additional snack betting on teams, or trading keys.

Unless I'm missing something, this is zero sum -- so it follows that a bunch of people mostly lost money (perhaps also during their childhoods)
hhmc
·10 bulan yang lalu·discuss
Sure theoretically. Outcomes are dramatically better if someone else is around
hhmc
·10 bulan yang lalu·discuss
I once did a course with a paramedic on basic aid. We were discussing choking, which is a condition that really needs a 2nd party to intervene. Someone asked what to do if you live alone (with no close neighbours) - the answer was essentially ‘good luck’
hhmc
·10 bulan yang lalu·discuss
Thanks, this is a fascinating note.

You’re right that really it’s collaborative production that’s the heart, and there’s no real benefit to LLMs.

It does feel like some classical explore/exploit algorithms could be interesting, but I imagine the challenge to any effort is really just getting enough high quality contributors
hhmc
·10 bulan yang lalu·discuss
I always thought there was decent promise in gwerns (gpt-3 era) proposal of a meta-llm CYOA: https://gwern.net/cyoa

Wonder if anyone ever took a crack at this
hhmc
·10 bulan yang lalu·discuss
Because the numbers are less than 1 sqrt pulls them towards 1 (the edge)