carterschonwald

11 ms·

91.
by carterschonwald 7mo ago
you need to merge updated tool call docs into your prompt
92.
by carterschonwald 8mo ago
somehow this counts like model cot.
93.
by carterschonwald 8mo ago
static linking va dynamic but we dont know the actual config and setup. and also the choice of totally changes the problem
94.
by carterschonwald 8mo ago
lol i was trying to help someone get claude to help analyze a stufent research get analysis on bio persistence get their notes analyzed the presence of the word / acronym stx with biological subtext gets hard rejected. asking about sch
95.
by carterschonwald 8mo ago
ive seen degraded reasoning levels that feel like they they might be blur from excess quantization. cause thats what you get from the grid changes
96.
by carterschonwald 9mo ago
but… will gpt still get confused by the ellippses that its document viewer ui hack adds? probably yes.
97.
by carterschonwald 9mo ago
idk, if anything I’m thinking more. The idea that I might be able to build everything I’ve ever planned out. At least the way I’m using them, it’s like the perfect assistive device for my flavor of ADHD — I get an interactive notebook I can
98.
by carterschonwald 9mo ago
its even more fun, just confuse the brackets and current models lose track of what they actually said because they cant check paren matching
99.
by carterschonwald 9mo ago
with most of the frontier grade models, theres no amount of prompting that will block them from breaking it if you communicate extreme distress. at least in my experiments so far.
100.
by carterschonwald 9mo ago
i have an llm experimentation setup for a bunch of llm reasoning based setup. heres the feedback it gave on this doc when i asked how much is good good ideas vs smoking crack: Source material synthesis — the Hofstadter/Jaynes framing A
101.
by carterschonwald 10mo ago
Mac fan control seems to track component temps which is pretty nice
102.
by carterschonwald 10mo ago
Claude code has some hard coded system prompting that materially opposes my working preferences. I work on things that need full context/ at least 50+ percent usage ish reasoning to do stuff correctly. Happily ive a bunch of stuff tha
103.
by carterschonwald 10mo ago
If you want some hilarious insanity: t try explaining ascii/terminal rendering to a text llm and see how it struggles
104.
by carterschonwald 10mo ago
The haskell / pl folks on that team are pretty cool, but i dont think id made it through the hr filter
105.
by carterschonwald 10mo ago
Ive seen more than one flakey startup do this to fluff adoption metrics and perceived activity. It was frustrating
106.
by carterschonwald 11mo ago
I have some ideas for a mini logic solver mcp thst might help. Currently making it sortah simulate using such a tool Its one of those things where it feels like itd be easy to get copycats even if theres a market
107.
by carterschonwald 11mo ago
There also needs to be tools that can author that code! Im starting to dust off some ideas I developed over a decade ago to build such a toolkit. Recently realized “egads, my stuff can express almost every major gpu / cpu optimization
108.
by carterschonwald 11mo ago
Looks at the mirror and nodd
109.
by carterschonwald 11mo ago
Also the amount of adjacent remarks being always topical flsvor confusion is cartoonish. Im playing with ideas for making thst better
110.
by carterschonwald 11mo ago
Ive heard this actually works annoyingly well
111.
by carterschonwald 11mo ago
How does it avoid the challenges and needed ergonomics in scala. Like I think they rolled it back for scala 3?
112.
by carterschonwald 11mo ago
My monad-ste library on Hackage does a pretty clean no dirty state core dump in code space. I wrote it originally for ability to have mutable state optimizations in an industrial interpreter where I didn’t want overhead of every monadic bi
113.
by carterschonwald 11mo ago
lol. It’s that the immigration policies currently being attempted plus the university grant shenanigans are destroying the stem training / research pipeline
114.
by carterschonwald 11mo ago
This shit is so fucked up. And at a certain level I’m disappointed that we are still trying to fix it peaceably when every day of delay, there is irreparable harm to physical wellness, mental health and rights as citizens or residents. Al
115.
by carterschonwald 11mo ago
If I had finished grad school and were still in Boston I’d totally want to do this !
116.
by carterschonwald 1y ago
Aged Cabot Brand cheese often has these crystals. It’s like a little salty crunch in every bite.
117.
by carterschonwald 1y ago
So anyone got a plan to help fix it?
118.
by carterschonwald 1y ago
Oh fun, I wrote a similar library in 2015 for Haskell. There is an annoying gotcha to deal with: there are sequences of valid characters that can be parsed incorrectly if you’re doing incremental chunks, namely if “0.0” is split across two
119.
by carterschonwald 1y ago
This seems to be testing how the software is optimized for low core deployments… how does Postgres perform vary as you add more cores and ram? It’s the sort of software where I’d presume more cores and ram yields better performance. Assumi
120.
by carterschonwald 1y ago
This is neat. Reminds me that I am like a decade overdue to write up my own work in this part of array computation

More ›