I love these tips from Jonathan Whitaker, just posted to answerdotai, on attacking 'high-surface-area problems'; i.e problems where 'when something doesn’t work it can be hard to find out where the issues may be, let alone what we need to do to fix them.'
answer.ai/posts/2024-04-…

some people get a great follow and think, 'i shouldn't say anything to stupid now, lest i shake them off.' other people think, 'lmao i'm going to double the fuck down this will be funny as hell.' the latter species of poaster is responsible for all cultural progress

I doubt that studying XZ will lead to any actionable improvement of security. Modern computing has a huge attack surface: No reasons the next bad actor would pick similar lines of action among almost infinite ones. So? More resources, less complexity, less dependencies.

You like nlp, huh? Name every word.

Probably if I had to define the most American 5 minutes of any movie it would be the “Life is a Highway” sequence in Cars

(guy training foundation model) just 0.1% MMLU away from product-market fit...

😭😭😭😭

all roads lead to rome

everything in data except SQL was a mistake

I’ve seen a lot written on ways to overcome impostor syndrome but I almost never see “keep writing, shipping, and asking for reviews on a lot of code” as a solution.

crossing out evernote and putting redis and looking back at the VCs for approval

hoping to avoid this unfortunate turn of events but every day I feel myself getting pulled closer to C or Rust