Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
lostmsu
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
13 ms
·
301.
▲
by
lostmsu
4mo ago
But you can not. At least not in US. Basically, you can not go to jail for it almost anywhere.
302.
▲
by
lostmsu
4mo ago
You almost convinced me, but then I remembered that you can sell that multi-million dollar home somewhere in the early 5-10 "years of grinding" and still end up the same.
303.
▲
by
lostmsu
4mo ago
You would be incorrect. That language model is decent in SVG pelicans and quite good at MIDI music.
304.
▲
by
lostmsu
4mo ago
Wow, that's very creative! /couldn't resist the pun/
305.
▲
by
lostmsu
4mo ago
The article stops short of actually telling why would anyone need NOT_CONSTANT
306.
▲
by
lostmsu
4mo ago
Compete? It is behind Kimi K2.6, which is in turn away behind Sonnet.
307.
▲
by
lostmsu
4mo ago
99% of the world would love lawyers, tax men, self-entitles software developers, and the likes to go the way of horse carriages.
308.
▲
by
lostmsu
4mo ago
OPUS is the default now, but there's more room. Especially with neural codecs.
309.
▲
by
lostmsu
4mo ago
This is a description of the life itself.
310.
▲
by
lostmsu
4mo ago
A video recording of a sample app would be nice
311.
▲
by
lostmsu
4mo ago
No video?
312.
▲
by
lostmsu
5mo ago
vllm-windows works well enough
313.
▲
by
lostmsu
5mo ago
Hopefully audio codecs will progress as well. I noticed that too. When I tried extreme screen recording compression with AV1 audio became a noticeable part of the bottleneck.
314.
▲
by
lostmsu
5mo ago
Same. I came back to do a little frontend work a couple of years ago and was horrified by the replacement of script tags with subresource integrity with npm and bundlers.
315.
▲
by
lostmsu
5mo ago
6MB should be enough for everyone!
316.
▲
by
lostmsu
5mo ago
The bottleneck in this setup is PCIe bus. You don't need optane to saturate it. 4 regular SSDs might do just fine.
317.
▲
by
lostmsu
5mo ago
> Rather, the author is just a really powerful and sophisticated pattern matching machine that only returns best guesses about what they “think” a question means. But the author doesn’t understand true meaning, not really. For all of the
318.
▲
by
lostmsu
5mo ago
Hm, Altman vs China. Peculiar choice.
319.
▲
by
lostmsu
5mo ago
You nailed it! There are established tools that handle the security model. MCP is the 5th leg.
320.
▲
by
lostmsu
5mo ago
You don't even need to train the model just to see if you can infer it at the claimed speed
321.
▲
by
lostmsu
5mo ago
I would say yes unless they are minors, but the laws in many places don't.
322.
▲
by
lostmsu
5mo ago
The question is: is it still worse than GPT 5.4?
323.
▲
by
lostmsu
5mo ago
If you open the page you will see what is reduced by how much.
324.
▲
by
lostmsu
5mo ago
The price cut is 50%.
325.
▲
by
lostmsu
5mo ago
The state of the art models (mostly GPT 5.5, but also Gemini and Claude) are better so they cost more. Qwen 3.7 Max is their only direct competition and it is not any cheaper.
326.
▲
by
lostmsu
5mo ago
2nd order effects are too complicated for average Joe
327.
▲
by
lostmsu
5mo ago
> it’s about having interesting stuff to do with your life This is the conclusion I came to as well. I do have kids. So I wholly agree with the sibling comment "compromise on lifestyle"
328.
▲
by
lostmsu
5mo ago
Canada counts it from the moment you entered it as domicile
329.
▲
by
lostmsu
5mo ago
You think 7 and 9 are fun, imagine the number is instead astronomically big as in BB(8).
330.
▲
by
lostmsu
5mo ago
This doesn't make sense. Both land sinking and water rising contribute together to 11mm/y ground level going closer to sea level. And the sinking has >2x effect.
More ›