Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
pr337h4m
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
91.
▲
I'm turning 41, but I don't feel like celebrating
(twitter.com)
72 points
by
pr337h4m
1y ago
|
123 comments
92.
▲
by
pr337h4m
1y ago
I don't think this can scale to really large models (300B+ params), especially once you add a little bit of RL for "common sense"/adversarial scenarios.
93.
▲
Memory leaks: the forgotten side of web performance (2022)
(nolanlawson.com)
2 points
by
pr337h4m
1y ago
|
0 comments
94.
▲
by
pr337h4m
1y ago
Come on, this is obviously a great thing. A lot of kids are gonna be super happy once this launches.
95.
▲
by
pr337h4m
1y ago
Now imagine they had a CBDC.
96.
▲
by
pr337h4m
1y ago
What do the OEMs have to say about this? A lot of them, including Samsung, have their own app stores. Surely they'd not be willing to cede control?
97.
▲
Announcing Tinker
(thinkingmachines.ai)
152 points
by
pr337h4m
1y ago
|
89 comments
98.
▲
by
pr337h4m
1y ago
soft porn
99.
▲
Writing a pre-mortem for an app that hasn't even been launched yet
(twitter.com)
1 points
by
pr337h4m
1y ago
|
0 comments
100.
▲
by
pr337h4m
1y ago
new funniest satoshi candidate just dropped
101.
▲
by
pr337h4m
1y ago
It will be unambiguously clear to history (maybe even in a decade) that SCOTUS's decision in TikTok, Inc. v. Garland was the worst ever since Korematsu.
102.
▲
Creating a decentralized bathroom (powered by the AT Protocol)
(dame.is)
2 points
by
pr337h4m
1y ago
|
0 comments
103.
▲
Who Is Elara Voss?
(maxread.substack.com)
3 points
by
pr337h4m
1y ago
|
0 comments
104.
▲
by
pr337h4m
1y ago
> I am unaware of any barrier, or any reason to infer there is one, for dextrous robots. Pretraining data?
105.
▲
"Middle-out compression" of LLM context on OpenRouter
(openrouter.ai)
3 points
by
pr337h4m
1y ago
|
0 comments
106.
▲
Nvidia and Intel to Develop AI Infrastructure and Personal Computing Products
(nvidianews.nvidia.com)
8 points
by
pr337h4m
1y ago
|
4 comments
107.
▲
Supplementary Information for the DeepSeek R1 paper [pdf]
(static-content.springer.com)
1 points
by
pr337h4m
1y ago
|
0 comments
108.
▲
How Opus and o3 saved me from permanent blindness
(mmaaz.ca)
14 points
by
pr337h4m
1y ago
|
6 comments
109.
▲
Oops, No Victims: The Largest Supply Chain Attack Stole 5 Cents
(securityalliance.org)
4 points
by
pr337h4m
1y ago
|
1 comments
110.
▲
Unsexy AI Failures: The PDF That Broke ChatGPT
(surgehq.ai)
1 points
by
pr337h4m
1y ago
|
0 comments
111.
▲
My Mom and Dr. DeepSeek
(restofworld.org)
13 points
by
pr337h4m
1y ago
|
0 comments
112.
▲
by
pr337h4m
1y ago
> The strategy we use relies on the referer header we see when we get an HTTP/S request. This is interesting data but is not really a useful estimate of search engine market share in 2025.
113.
▲
LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videos
(linjohnss.github.io)
5 points
by
pr337h4m
1y ago
|
0 comments
114.
▲
The Machinery of Freedom [pdf]
(daviddfriedman.com)
3 points
by
pr337h4m
1y ago
|
0 comments
115.
▲
by
pr337h4m
1y ago
Never take government money. RIP Intel.
116.
▲
Claude says “You're absolutely right!” about everything
(github.com)
773 points
by
pr337h4m
1y ago
|
533 comments
117.
▲
How attention sinks keep language models stable
(hanlab.mit.edu)
219 points
by
pr337h4m
1y ago
|
36 comments
118.
▲
by
pr337h4m
1y ago
In this case, compromise is equivalent to surrender
119.
▲
by
pr337h4m
1y ago
This could be repurposed for really good stylistic search for music (similar to how CLIP powered stylistic search for images on same.energy)
120.
▲
by
pr337h4m
1y ago
Related: https://vgel.me/posts/representation-engineering/ https://github.com/vgel/repeng
More ›