Sam Foley
@physfoley.bsky.social
Physics of membranes and self-assembly || Biophysics postdoc at Johns Hopkins || Previously Penn State -> Carnegie Mellon || for I perplex others not because I am clear, but because I am utterly perplexed myself ||
www.samuelfoley.com
www.samuelfoley.com
So everyone knows the pickleback, but what about...
November 27, 2025 at 2:50 AM
So everyone knows the pickleback, but what about...
Benefit of having an L-desk at home: desk buddy
November 26, 2025 at 8:28 PM
Benefit of having an L-desk at home: desk buddy
Todd finally got to me. Bought Skyrim for a second time this weekend. At least I'll get to play the DLCs for the first time.
November 24, 2025 at 5:30 PM
Todd finally got to me. Bought Skyrim for a second time this weekend. At least I'll get to play the DLCs for the first time.
"Their words cannot be allowed to stand"
November 20, 2025 at 4:48 PM
"Their words cannot be allowed to stand"
Favorite soft drink of a computational physicist? LSODA
November 19, 2025 at 10:17 PM
Favorite soft drink of a computational physicist? LSODA
Genuinely incredible reddit comment on why C++ is called "C++" and not "++C"
November 17, 2025 at 7:29 PM
Genuinely incredible reddit comment on why C++ is called "C++" and not "++C"
Reads like this classic
November 17, 2025 at 2:27 PM
Reads like this classic
First, we have to pick a fish. A quick google says a Salmon weighs roughly 5kg. I am going to make the crude approximation that the fish is 100% water, and we want to completely boil that water into steam.
November 10, 2025 at 8:23 PM
First, we have to pick a fish. A quick google says a Salmon weighs roughly 5kg. I am going to make the crude approximation that the fish is 100% water, and we want to completely boil that water into steam.
A question came up in the lab today:
How many fish could you (completely) boil with the energy it takes to train an LLM like ChatGPT?
How many fish could you (completely) boil with the energy it takes to train an LLM like ChatGPT?
November 10, 2025 at 8:23 PM
A question came up in the lab today:
How many fish could you (completely) boil with the energy it takes to train an LLM like ChatGPT?
How many fish could you (completely) boil with the energy it takes to train an LLM like ChatGPT?
"Why do you like membrane bending simulations so much?"
November 6, 2025 at 5:42 PM
"Why do you like membrane bending simulations so much?"
Wow, I had never heard of marimo before, but if their FAQ is for real, it solves the #1 complaint I've had about notebooks for years! (and the main reason I've recommended inexperienced programmers not to use them)
October 24, 2025 at 8:25 PM
Wow, I had never heard of marimo before, but if their FAQ is for real, it solves the #1 complaint I've had about notebooks for years! (and the main reason I've recommended inexperienced programmers not to use them)
Always good to remember that even the most advanced LLM AI models currently available have extremely high rates of hallucination, and all outputs should be thoroughly validated.
source: research.aimultiple.com/ai-hallucina...
source: research.aimultiple.com/ai-hallucina...
October 21, 2025 at 5:03 PM
Always good to remember that even the most advanced LLM AI models currently available have extremely high rates of hallucination, and all outputs should be thoroughly validated.
source: research.aimultiple.com/ai-hallucina...
source: research.aimultiple.com/ai-hallucina...
Wow my gf is cuddling with the actual Batman rn
October 21, 2025 at 12:02 AM
Wow my gf is cuddling with the actual Batman rn
Here's something I haven't seen before: Google search apparently doesn't just directly feed your query to its "AI", but can try to distill your query into a shorter input, which in this case leads to a confused answer from the LLM. Trying to reduce token count to save on computation costs?
October 13, 2025 at 4:30 PM
Here's something I haven't seen before: Google search apparently doesn't just directly feed your query to its "AI", but can try to distill your query into a shorter input, which in this case leads to a confused answer from the LLM. Trying to reduce token count to save on computation costs?
POV: laying on the couch with kitty cat
October 9, 2025 at 2:24 AM
POV: laying on the couch with kitty cat
Leaves the door open for future awards, I suppose
October 7, 2025 at 5:23 PM
Leaves the door open for future awards, I suppose
A very expensive day later, and the consensus seems to be: he had a tummy ache
October 3, 2025 at 10:32 PM
A very expensive day later, and the consensus seems to be: he had a tummy ache
I know there are more important things to worry about, but the NIH website main banner being a super shitty AI-generated image pisses me off so much. "Know your blood pressure numbers" yeah they got me there, that is good advice after seeing the picture.
October 1, 2025 at 4:51 PM
I know there are more important things to worry about, but the NIH website main banner being a super shitty AI-generated image pisses me off so much. "Know your blood pressure numbers" yeah they got me there, that is good advice after seeing the picture.
"Nvidia to invest $100 billion into OpenAI"
September 22, 2025 at 4:50 PM
"Nvidia to invest $100 billion into OpenAI"
Hard mode: "draw me an ASCII art axolotl"
The result:
The result:
September 19, 2025 at 3:28 PM
Hard mode: "draw me an ASCII art axolotl"
The result:
The result:
The small one almost works, but then... what even is that second abomination?
September 19, 2025 at 3:26 PM
The small one almost works, but then... what even is that second abomination?