kcb

9 ms·

91.
by kcb 10mo ago
I don’t care if the whole game from end to end is generative AI if it’s an incredible game. Having a moral stance against a specific use of floating point numbers and algorithms in a medium filled with floating point numbers and algorithms
92.
by kcb 10mo ago
The Thinkpad x13s is pretty good for a mostly normal arm Linux laptop. The SOC is still pretty usable and it's functionality on Linux is also pretty complete. The main thing lacking is the speakers are very low to protect them from blo
93.
by kcb 10mo ago
That's not very encouraging because that statement has been true most every day in computing for the past 50 years. The rate at which the compute per dollar increases is what matters.
94.
by kcb 10mo ago
Heh, I was looking at an eks cluster recently that was using Cast AI autoscalar. Scratching my head as there was a bunch of inf instances. Then I realized it must be cheap spot pricing.
95.
by kcb 10mo ago
LLMs require memory and interconnect bandwidth so needs a whole package that is capable of feeding data to the compute. Crypto is 100% compute bound. Crypto is a trivially parallelized application that runs the same calculation over N input
96.
by kcb 10mo ago
iPhone docks and speakers were already obsolete. They had a moment during the 30-pin era, but its been long since Bluetooth, Carplay took over in any mainstream use.
97.
by kcb 11mo ago
Can also use a micro SD adapter through the mem card slot. Generally considered a bit better experience than a USB drive.
98.
by kcb 11mo ago
Not too hard. The original run command is stored if you inspect a running container.
99.
by kcb 11mo ago
Yea, there's a Bluetooth protocol for it called OBEX.
100.
by kcb 11mo ago
That's not the case, CUDA containers user space does not have to match the host drivers CUDA capability. The container needs to be the same major version or lower. So a system with a CUDA 13 capable driver should be able to run all pre
101.
by kcb 11mo ago
TSMC in Arizona has been at full scale production of 4nm chips for a while now.
102.
by kcb 11mo ago
What are you talking about? This change for "experienced users" was only just announced and not part of any previous announcement. It has not been clear from the start at all.
103.
by kcb 11mo ago
> Is there anywhere within reasonable commute radius left to develop around NYC at scale? Many many places, if we were willing to undertake large scale public transit projects again. But no one even bothers to run on that anymore.
104.
by kcb 11mo ago
They'll just need to start from scratch on all this. https://substackcdn.com/image/fetch/$s_!CwyW!,f_auto,q_auto:... I don't think some power efficient laptop SoCs gives you much competitive advantage th
105.
by kcb 1y ago
No shot are you doing tens of millions of anything useful per second per core. That's like beyond HFT numbers.
106.
AMD Solarflare X4 NICs Launched for Low Latency Trading(servethehome.com)
3 points by kcb 1y ago|0 comments
107.
by kcb 1y ago
If we're talking space colonies that house large numbers of people wouldn't they be large enough to spin for artificial gravity?
108.
by kcb 1y ago
Shame to still see newly released products using a 13 year old core design. How has there been such little progress on low power ARM cores that it still makes sense to build a Cortex-A53 based soc on a modern node.
109.
by kcb 1y ago
Local as in datacenter is the key there. The original comment was about end user devices.
110.
by kcb 1y ago
30k in a month is an enormous amount of tokens with Claude through AWS Bedrock. And companies already commonly trust AWS with their most sensitive data.
111.
by kcb 1y ago
What's the benefit to running LLMs locally? Data is already remote, LLM inferencing isn't particularly constrained by Internet latency. So you get worse models, performance, and battery life. Local compute on a power constrained m
112.
by kcb 1y ago
Remember there are likely Google developers on this site. Maybe some of the same ones that are implementing this change.
113.
by kcb 1y ago
The difference is crypto wasn't memory and throughput dependent. That's why a small asic on a USB stick could outperform a GPU.
114.
by kcb 1y ago
I don't think they have to do that for a Steam Linux release. Steam has Linux native runtimes to provide games with a consistent environment across distros. https://github.com/ValveSoftware/steam-runtime/blob&
115.
by kcb 1y ago
Americans are good at spending money.
116.
by kcb 1y ago
Especially a product that demands access to large quantities of your most sensitive data to be useful.
117.
by kcb 1y ago
I've used nv-ingest and Nvidia's nemoretriever-parse model.
118.
by kcb 1y ago
I have a 5070 ti running Kubuntu 25.04 and it's a mess. Animations repeating, half the desktop disappears when waking from sleep, HDMI audio cuts out... I swapped to a 7900xt and it is absolutely flawless.
119.
by kcb 1y ago
Kernel driver only.
120.
by kcb 1y ago
But my Grandma might install the wrong extensions. Please think of her.

More ›