winwang

5 ms·

31.
by winwang 4mo ago
Minor note, 2x $/tok is not 2x cost. Personally, I see Fable being significantly more token-efficient than Opus 4.8. Then, there's also the compounding costs of quality.
32.
by winwang 4mo ago
My experience has been that 5.4 is slower than 5.5 (confound: I use >512k max context size for 5.4, though it seems slower even below the normal size)
33.
by winwang 4mo ago
I typically just launch CC with `--model claude-opus-4-6[1m]`, `4-6[1m]` -> `4-8[1m]` works fine. Still 200k max without the `[1m]`.
34.
by winwang 4mo ago
There's the other (orthogonal) possible explanation of using more GPUs for stress-testing before product launch.
35.
by winwang 4mo ago
How else would you write this (marketing copy) exactly? "Its output matches better to its CoT which matches to better to our hidden state decoder according to <insert measure here>; see <insert paper ref>"? ... Actuall
36.
by winwang 4mo ago
Awesome, thanks for posting because I think I hit a possibly-spurious bug in turning Adaptive off when I switched models (4.6 -> 4.8, extra). Tried again, works as intended (I hope). More importantly for me, though, is how CC will respon
37.
by winwang 4mo ago
Let's hope I don't have to disable it after a day like with 4.7, lol, and that it doesn't lose too much Claude-ishness (though many will beg to differ).
38.
by winwang 4mo ago
Yes, but that's also a specific luxury I can choose for myself. Definitely a fun and interesting question. At some level of reliance, people would answer "no", but there's the large middle ground (assuming similarly-fron
39.
by winwang 5mo ago
Yep. No one bats an eye at eyewitnesses "hallucinating" details, or that I'd rather have Opus as a coworker vs a random middle schooler (err, labor laws notwithstanding). I think perhaps too much of the dialogue around intell
40.
by winwang 5mo ago
This is, ironically, a pretty good idea. ...Minus the fact that you're presumably talking about having AI generate it all instead.
41.
by winwang 5mo ago
I think most people agree with you -- that's why. Also because I'd say most programmers don't care much about maintainability or quality. I personally find that AI writes better Scala than Python.
42.
by winwang 5mo ago
Yeah, I pretty much agree. Opus and GPT will both come up with the most "organically-grown" "designs" if you let them. They do slightly better when asked to design first, but they seem to avoid many important questions (
43.
by winwang 5mo ago
I would presume this is shorthand for something like "generated text which would normally be classified as belief". I guess a more ridiculous response could be "what does it mean for a miserable pile of secrets to believe som
44.
by winwang 5mo ago
I absolutely feel like a "different" part of my mind is loaded when seriously engineering something myself vs vibecoding+reviewing. Even the reviewing is more annoying in the latter mental context.
45.
by winwang 5mo ago
Honestly, I gotta agree, I find that I get way more frustrated with Claude recently than Codex.
46.
by winwang 6mo ago
Obviously nowhere near Erdos problem complexity but I've been using GPT (in Codex) to prove a couple theorems (for algos) and I've found it a bit better than Claude (Code) in this aspect.
47.
by winwang 6mo ago
That's usually not how these things work. Only parts of the prompt are actually loaded at any given moment. For example, "system prompt" warnings about intellectual property are effectively alerts that the model gets. ...Thou
48.
by winwang 6mo ago
Only somewhat related but there is supposedly a SIMD/GPU-friendly skiplist algo written about here: https://csaws.cs.technion.ac.il/~erez/Papers/GPUSkiplist.pdf
49.
by winwang 6mo ago
Each SM should have 4 independent SMSPs (32 lanes each), no? Effectively a "4-core" task-parallel system per SM.
50.
McGridsort: Warping Grids for GPU k-way mergesort(winwang.blog)
3 points by winwang 6mo ago|1 comments
51.
by winwang 6mo ago
Had a fun little idea for a weird GPU/SIMD k-way mergesort a couple years back, finally decided to write it up! (Anti-)jumpscare: no hard perf numbers in the post (though I have profiled it somewhat already).
52.
by winwang 7mo ago
Interestingly, I find that the models generalize decently well as long as the "training" (more analogous to that for humans) fits in (small enough) context. That's to say, "in-context learning" seems good enough for
53.
by winwang 7mo ago
How much of this is expectations setting by the heights models reach? i.e. of we could assess a consistent floor of model performance in a vacuum, would we say it's better at "AGI" than the bottom 0.1% of humans?
54.
by winwang 7mo ago
It would be much worse if it had said "You are absolutely wrong to be confused", haha.
55.
Non-Messing-Up++: Diagonal Sorting and Young Tableaux(winwang.blog)
14 points by winwang 7mo ago|1 comments
56.
by winwang 7mo ago
Hey HN, I figured to just share this for feedback despite its dry-ness and small-idea-ness.
57.
by winwang 7mo ago
(no idea but) I feel like changing the first number has a psychological issue, but the 2nd number feels more important than just "minor" sometimes. So may as well let the schema set the mind free?
58.
by winwang 7mo ago
...I almost thought it was a parody site!
59.
by winwang 7mo ago
Interesting. I've felt like it's never been easier to learn things, but I suppose that's not quite the same as "acquiring new skills". I don't know if it applies, but it's always been easy to take the easy
60.
by winwang 7mo ago
Yeah that's somewhat close to what I meant, though there's an irony here in that your comment (and this one) are pretty reddit-esque.

More ›