Wil Gieseler
banner
wilgieseler.com
Wil Gieseler
@wilgieseler.com
just tryna be good https://wilgieseler.com/
November 9, 2025 at 9:37 PM
mood
November 9, 2025 at 9:28 PM
persons per bedroom has been declining www.huduser.gov/publications...
November 9, 2025 at 9:23 PM
in the 90s everyone could afford an apartment like this in NYC. i saw it on television!
November 9, 2025 at 8:38 PM
November 9, 2025 at 6:47 PM
you gotta scroll!
November 8, 2025 at 2:21 AM
November 7, 2025 at 4:22 AM
November 7, 2025 at 4:21 AM
i know you know about inflation but it’s fun to consider whether something like this actually did get more expensive or if it’s just a mirage from the numbers getting higher. this loaf is the same real value as a $0.15 loaf in 1900. it’s possible it’s been getting cheaper!
November 1, 2025 at 7:30 PM
that's fine, but the question is does it exceed maintenance? your entire thing here relies on the maintenance somehow blowing up the cost story because its a unique vehicle, but i don't see why that would be the case, and they could also engineer it for low maintenance in taxi use cases
October 29, 2025 at 3:34 AM
POV: you're about to not laugh
October 19, 2025 at 10:54 PM
#nokings santa monica
October 18, 2025 at 8:50 PM
touch grass
October 16, 2025 at 10:27 PM
October 14, 2025 at 1:36 AM
October 14, 2025 at 1:33 AM
October 14, 2025 at 1:26 AM
not so, one guy is locked in
October 6, 2025 at 10:28 PM
funniest disclaimer i have ever seen
September 23, 2025 at 5:00 AM
Great call. Cancel Disney+ immediately and don't spend money at Disney. Right now a final decision is pending. Now is the best time.
September 18, 2025 at 5:35 PM
call it whatever you want. knowing its own capabilities is also a difficult thing to get it to do. thats just how the engineering works out. that is one of the major metrics they are focusing on.
August 8, 2025 at 3:20 AM
the model only gets the numbers, so analyzing the internal contents of words is specifically difficult. the thinking models work around this. but it means that this particular task is unusually ill-suited to be performed directly by an LLM
August 8, 2025 at 2:43 AM
they just always use the non thinking model for this specific question. this question is essentially a demonstration of why the thinking models are useful

(because the LLM does not see the letters that go into it directly, they are converted to tokens or chunks of words so it never sees the text)
August 8, 2025 at 2:41 AM
the models need to "think" to get this type of question right, because of the tokenization problem which thinking can mitigate
August 8, 2025 at 2:39 AM
the models need to "think" to get this type of question right, because of the tokenization problem which thinking can mitigate
August 8, 2025 at 2:38 AM
August 6, 2025 at 2:42 AM