great day to push some breaking changes to prod. Get it done, folks.

imagine if civil engineers had the same urge to just rebuild from scratch instead fix a few bugs

This is basically what Bayesian inference is

Well at first I thought I was far too masculine to literally use soap, but this name completely changed my mind

this is phd energy

brb gotta go network with my 28,999 closest colleagues

for my next trick, i will look at the data

I'm admittedly a mac kinda guy, but it is funny that every windows PC i've owned has like 7 petabytes in the downloads folder and every macbook's memory is 98% full the first time i boot it up

Wow, what a year.

lol do you guys remember smoking sections of restaurants? What a bonkers concept in retrospect

Reminding myself this every 17 minutes or so today

Gotta be honest, this is the first time ChatGPT has got an earnest laugh out of me.

This holiday I hope my colleagues are thankful for all the PRs I’m making today.

I will immediately switch to any music streaming service that provides a "don't consider children's music when suggesting music" option. All my suggested playlists are just mickey mouse and elmo.

I liked grep before it was cool

I name all my devices after Taco Bell menu items. I am currently typing this from Grilled Stuft Nacho.

It’s kind of crazy that you can pinpoint the exact year everyone collectively put their phones on silent by looking at the Apple Ringtones store.

gpt-5.1.3a-codex-really-max-high-fast-zoom-zoom

It’s probably a moot point because toon is just spicy CSV, but say you came up with the most token efficient way to represent data possible - would it even be that useful if it wasn’t in the LLM’s training set? Feels like it’d hurt performance

It’s kind of funny that we had like 2 years of “hallucinations are unacceptable” discourse and then hard pivoted to a bunch of “let me get you a wrong answer faster” agentic coding models.