seen from Malaysia
seen from China
seen from Australia
seen from United States
seen from Canada
seen from United States
seen from Philippines
seen from Australia
seen from China
seen from China
seen from China
seen from United States
seen from Taiwan
seen from United States
seen from United States
seen from United States
seen from China
seen from Canada
seen from United States
seen from China
impressed that github copilot can now generate incorrect regexes much faster than i can, and argue for their validity more persuasively
(i jest - it got it right the first time, i think, but then when i asked whether two of the patterns were redundant, it incorrectly told me that they were. and patted me on the head for being such a clever programmer to notice this)
pleased to report i solved a programming issue which claude 4 opus repeatedly failed on. i might* avoid the glue factory another couple years.
*the reader will charitably ignore that the problem was created by my own poor function design choices
*dracula flow voice* i'm getting xgboost errors with 0 results on the repo. my cuda stack trace look like the tower of babel
i think a lot of the productivity gains from LLM coding tools are from people who weren't previously using linter, autocompletion (with appropriate type hinting in dynamically typed languages), or even the debugger. the LLM can kind of substitute for those tools, and has a lower barrier to entry. but watch out
NUMAmusume. horsegirls in local cache are faster than more remote horsegirls. is this anything
at the biweekly team check-in where we describe our current mental state with a percentage value
me, thinking: i'm super butthurt about how the last week has gone, how i worked over the weekend to make our product worse just to appease the allmighty deadline. but obviously i can't rate my well-being lower than the iranian coworker, or the guy whose wife died last month. that would be an unrecoverable social disaster. maybe a 75?
dead wife guy: so i'm probably at a 75, got to hang out with some friends this weekend, and the weather is really improving . . .
me, thinking: ok i can work with this. 80%? can i get away with an 80% here?
iranian guy: so i am at a 97%
import optuna
optuna.trial.suggest_forcefem()
good software engineering practices don't always take into account the demands of day-to-day experience in the industry. for example, i tried adding a unit test to the new function i was adding, but i ended up having to spend a bunch of time debugging it because half the assertions were failing
