Gal Sapir

@sapir.bsky.social

66 Followers 302 Following

https://sparsethought.com/

example of such internal thoughts
RCQ scores disaggregated by evaluation dimension; sample sizes as in c. e, Refusal rate (n = 100 questions per model)
General-purpose large language models outperform specialized clinical AI tools on medical benchmarks
screenshot of the console showing a lot of token usage today
screenshot showing codex ran for >3 hours
My eLetter
claude code suggesting to clear when you come back to a session after a while
a screenshot showing claude for desktop saving memories
copy as markdown - seen on the gh GUI
a screenshot from the marginalia web app
Karri's tweet about the issue of "focus loss" - people lose focus and build things that don't promote the original goal, effectively lost in side quests
A diagram created with the tool
running more than one subagent
התכנית
סמסונג מה עשיתם עם הפוקוס בתמונה הזאת
Hellolulu