Catloafdev

5 ms·

91.
by Catloafdev 3mo ago
Ah ya then that's a bit of a gap. For anyone considering these devices, the only reason I would recommend against them is if you plan on getting multiple to link together - the DGX Spark has a much, much faster interconnect bandwidth c
92.
by Catloafdev 3mo ago
These devices were great when they were cheaper than the DGX Spark. But when they cost the same price (unless the Spark has shot up too), there's no reason to buy this over a Spark. The Spark is literally a faster version of this, with
93.
by Catloafdev 3mo ago
Is it really that expensive to not do stuff like this? I guess 'turning the entirety of the American public against data centers' is not something they factor into the cost analysis.
94.
by Catloafdev 3mo ago
It depends - for what? If your security model is sandboxing an agent to ensure they don't nuke your PC, then there are a lot of options, you can use something like bubblewrap[1] or a microVM like libkrun[2] if your goal is light-weight
95.
by Catloafdev 3mo ago
I don't see any info about what laws or actions specifically are happening. Is there more info somewhere?
96.
by Catloafdev 3mo ago
Apologies, I appreciate you posting this!
97.
by Catloafdev 3mo ago
They serve two different purposes. Edited*
98.
by Catloafdev 3mo ago
Closed date: 02/23/2026
99.
by Catloafdev 3mo ago
Conceptually, sure - but crypto? Really?
100.
by Catloafdev 3mo ago
DDoS protection and the number of features they offer are kind of unmatched. I often see threads complaining about Cloudflare, never see suggestions for better alternatives.
101.
by Catloafdev 3mo ago
This feels like a 'Horse Armor' moment. I expect much more of this type of thing going forward.
102.
by Catloafdev 3mo ago
I don't have a 'source' off-hand but I recommend reading up on it if you want to learn more. A lot of models on HF show a card demonstrating the different quality trade-offs between quants.
103.
by Catloafdev 3mo ago
Q8 is virtually lossless. The quantization is much more noticeable around Q4 and below. FP16->Q8 on consumer hardware is 2x the speed at ~99.99% the quality.
104.
by Catloafdev 3mo ago
Nobody runs unquantized, there's literally no reason to. Q8 would be the largest anyone actually runs on consumer hardware for inference.
105.
by Catloafdev 3mo ago
The model they reference can be easily run with 24gb+ of VRAM, and there are other similar models capable of running easily on 16gb of VRAM. It's not like 128gb is a requirement here.
106.
by Catloafdev 3mo ago
Hello, it's the internet calling, today is that day. https://github.com/ikawrakow/ik_llama.cpp Edit: it's gonna be slow if you're not using any VRAM. But it's possible. Software isn't going to
107.
by Catloafdev 3mo ago
I don't see how this can really go anywhere, but it'd be nice if it does. Not like it'd be the first time someone shook a stick at them. At this point the only hope for change is if China finally decides to get in the game ra
108.
by Catloafdev 3mo ago
It'd be cool to see more of this type of thing, but I have to imagine the ability for it to be updated to a brand-new model as new models come out is limited. If that is the case, it's going to be an extremely hard sell.
109.
by Catloafdev 3mo ago
I'd be shocked if it was anything more than this.
110.
by Catloafdev 3mo ago
Image models need software like ComfyUI, not LM Studio.
111.
by Catloafdev 3mo ago
Because it's not about consumer lock-in, it's about enterprise lock-in. That's what they're chasing. Regular users are basically marketing.
112.
by Catloafdev 3mo ago
Affordability is not the current goal. Vendor lock-in is the current goal. Consumer prices are a drop in the bucket comparatively.
113.
by Catloafdev 3mo ago
You think having a document detailing competitors is the same thing as compiling personal information of people who have publicly commented against what you're doing? The sandbagging on this story is crazy.
114.
by Catloafdev 3mo ago
I think people are going to continue to be surprised by the capability of small models. Now, if you ask this model to have a conversation with you, it's gonna fail and be incoherent. But boy, does it sure reason through math problems w
115.
by Catloafdev 3mo ago
3.6 scores better on coding across the board. Edit: specifically Qwen 3.6 27B beats that on coding and agentic workflows.
116.
by Catloafdev 3mo ago
Crazy to see this attempt to be normalized here. No. No, this is not normal.
117.
by Catloafdev 3mo ago
I think the reality is at this point the frontier regards CoT as extremely valuable, none of them are giving you genuine CoT anymore. I don't think there is any future in attempting to measure or evaluate CoT from frontier models - I e
118.
by Catloafdev 3mo ago
Oh I should have clarified, I meant 'in the context of releasing a public tool' I get that both need to exist as tools. I just don't see any safe way of doing a truly public release of the offensive end of it, you'd need
119.
by Catloafdev 3mo ago
> Owning money should allow you more representation than not owning money The capitalist's manifesto
120.
by Catloafdev 3mo ago
> Within the next two years you’ll stream a full feature film generated on demand (your mood, your cast, your language) for the cost of a search query Tapped out on this line, can't take this seriously.

More ›