Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
KaoruAoiShiho
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
61.
▲
by
KaoruAoiShiho
1y ago
Nah AI can easily be programmed to be much more patient and investigate edge cases and figure out personalized solutions thoroughly and provide bespoke service. This problem would be solved, though of course there are other issues with bias
62.
▲
by
KaoruAoiShiho
1y ago
Yep this is something that only AI can solve. Same situation applies to education, sales, HR. Human powered bureaucracies and systems suck.
63.
▲
by
KaoruAoiShiho
1y ago
I hate to say this but when I saw this line: > My continued shareholding isn't just a matter of financial confidence—it's a statement of faith in what HP can become when the right leadership applies systematic thinking to innov
64.
▲
by
KaoruAoiShiho
1y ago
They already have a chatgpt competitor, you can talk to llama in all the facebook apps, they're just clearly inferior.
65.
▲
by
KaoruAoiShiho
1y ago
This is a new organization, Lecun leads the FAIR team which is more focused on basic ML research than this ASI focused team.
66.
▲
by
KaoruAoiShiho
1y ago
Seems like a strawman... I don't know anyone who thinks they "will be on top" and beat the AI, it's more of a let's just make as much money as we can for now before AGI ruins us all.
67.
▲
by
KaoruAoiShiho
1y ago
The Sims, Crusasder Kings, the new inzoi games all have this.
68.
▲
by
KaoruAoiShiho
1y ago
You're right. I think it's extremely cultural to Christianity. The notion of suffering for your sins, and when bad things happen you ought to try to feel it as much as possible instead of working on pain mitigation. This is defini
69.
▲
by
KaoruAoiShiho
1y ago
The conflicts in his plays just aren't that profound, you know? Compared to modern "psychological turmoil" it's relatively simplistic, something like Macbeth's "ambition vs morality". Modern writers pull i
70.
▲
by
KaoruAoiShiho
1y ago
If he did it's not shown in his writing, or really any pre-20th century writers really.
71.
▲
by
KaoruAoiShiho
1y ago
I recommended a book in my comment here: https://news.ycombinator.com/item?id=44181274
72.
▲
by
KaoruAoiShiho
1y ago
depth and characterization isn't everything but it's the discussion chosen by the OP, he seemed to imply that Dostoevsky had the best psychological interiority and "multiple scales of experience".
73.
▲
by
KaoruAoiShiho
1y ago
I might've conveyed myself poorly. Never said Dostoevsky wasn't great or not a relevant read! I was just reacting to the OP who seemed to say it's still the best ever. But we've improved everything, art, kites, the wheel
74.
▲
by
KaoruAoiShiho
1y ago
I did 30 novels a year in highschool, it's nothing crazy. Though tbf kids do have more free time than working adults, I think it would really only drop if you become parents though.
75.
▲
by
KaoruAoiShiho
1y ago
Lol I always laugh when I see people talk up Dostoevsky and any 19th century writer. It's 2025 now and literature has improved in depth and characterization as much as any other field.
76.
▲
by
KaoruAoiShiho
1y ago
How does it compare to the other recent YC company Stack Auth?
77.
▲
by
KaoruAoiShiho
1y ago
Does this work on any language or text?
78.
▲
by
KaoruAoiShiho
1y ago
Aider is not just leetcode exercises I think? livecodebench is leetcode exercises though.
79.
▲
by
KaoruAoiShiho
1y ago
Is this really worthy of a claude 4 label? Was there a new pre-training run? Cause this feels like 3.8... only swe went up significantly, and that as we all understand by now is done by cramming on specific post training data and doesn'
80.
▲
by
KaoruAoiShiho
1y ago
iPhone should've been a successful formfactor, it's not the form factor that's the problem it's the lack of apple datacenters that can train and inference for a popular service. Instead Apple went all in on edge inferenc
81.
▲
by
KaoruAoiShiho
2y ago
That describes Google buying Youtube and Facebook buying WhatsApp, those seem to have turned out okay.
82.
▲
Benchmarking LLMs over Long Context
(fiction.live)
1 points
by
KaoruAoiShiho
2y ago
|
0 comments
83.
▲
by
KaoruAoiShiho
2y ago
It's much higher than 1.5x-2x, both scale up and scale out. Moving onto fp4 alone will offer a huge speedup.
84.
▲
by
KaoruAoiShiho
2y ago
Dario, founder of Anthropic is an ex-Baidu AI employee, it was at Baidu that he learned the bitter lesson.
85.
▲
by
KaoruAoiShiho
2y ago
The TikTok ban is still happening, some trump donator will end up owning majority of tiktok.
86.
▲
by
KaoruAoiShiho
2y ago
How good is the RAG though? Just throwing a vector db at it doesn't make it useful...
87.
▲
by
KaoruAoiShiho
2y ago
Would be interested in using this if not for the lack of LLM support. Can you also explain what the advantages are over React Native?
88.
▲
by
KaoruAoiShiho
2y ago
It's necessary to defend search, even if this cannibalizes search it's better they do it than let someone else do it.
89.
▲
by
KaoruAoiShiho
2y ago
Dropped could also mean they used to use it but stopped, that's also pretty common parlance in software...
90.
▲
by
KaoruAoiShiho
2y ago
Everything messes with your head, this adjusts the mess levels.
More ›