kzrdude

6 ms·

151.
by kzrdude 3mo ago
A common factor seems rather that international sports orgs are often very corrupt. Would be interesting to understand why it happens systematically. Lack of good governance structures must be part of it.
152.
by kzrdude 3mo ago
How it works in practice is that some release flows add wheels for different platforms, as they get ready, separately.
153.
by kzrdude 3mo ago
Part of the problem is that features are not added fully This here should not need to happen: > document that moved-from objects cannot be used It's moved from, so of course it cannot be used. Shouldn't have to add an assertion
154.
by kzrdude 3mo ago
Go is great if you prefer deepseek V4 flash. Then it's very hard to use up the allowance.
155.
by kzrdude 3mo ago
Maybe this is relevant? Just in case > For autonomous subagents with tool calls, code execution, or multi-step reasoning: set thinking_level to "medium" or "high" to prevent premature tool termination. I just happened
156.
by kzrdude 3mo ago
He's already proven repeatedly that he'll break out of containment
157.
by kzrdude 3mo ago
It's smarter than DeepSeek v4 Pro (preview) on several benchmarks, like HLE.
158.
by kzrdude 3mo ago
Well, you can use the google models from Pi. Go to aistudio.google.com and set up an API key. There is a free quota, it's relatively large for the Flash Lite models. ...and last time I looked the limits were more generous for Gemma 4 t
159.
by kzrdude 3mo ago
I asked Gemma 4 E2B, and if you use it as a reasoning model, it will give a better answer (that it doesn't know; it was also using the date information from the prompt.)
160.
by kzrdude 3mo ago
Google's models are always very well spoken and much more pleasant to talk to. And I've used the open weight models a lot, too.
161.
by kzrdude 3mo ago
I've never questioned the word lite before because it's existed my whole life.. So does it make sense? Why does it exist and where does it come from? More than coming from "light".
162.
by kzrdude 3mo ago
That sounds smart, maybe worth testing, but I think it bloats the instruction cache even when not tracing.
163.
by kzrdude 3mo ago
If you go small enough it should be no problem. For example Gemma 4 E4B in Q6 or Q4 quantization should run well on your laptop. It shouldn't be too taxing, but would still want to eat 7-9 GB of VRAM or so. Now that model is mostly use
164.
by kzrdude 3mo ago
LM Studio seems to do the same thing, except that LM Studio is not open source. So they have a point, they do something more.
165.
by kzrdude 3mo ago
Looks like Portuguese to me (I don't speak it)
166.
by kzrdude 3mo ago
Wikipedia has policies and it needs to use reliable sources. This rule is often skirted and a lot of facts are cited to self-published and fast moving web sources. Those who are braking the progress on the article here are doing the right t
167.
by kzrdude 3mo ago
Which exact model does claude use here?
168.
by kzrdude 3mo ago
Mario Zechner's talk "Building Pi in a World of Slop" is a good listen, on the topic of adults in a "world of slop". https://www.youtube.com/watch?v=RjfbvDXpFls
169.
by kzrdude 3mo ago
Kimi K3 is on the opencode-go plan's price list now. I haven't confirmed it, but I guess that means it's available. (Not in Zen.)
170.
by kzrdude 3mo ago
Fwiw, American industry has given away a lot for free - you could include large parts of the open source movement in that - and all the "free" VC backed services like facebook would be another prong of the same comparison. I would
171.
by kzrdude 3mo ago
If DeepSeek v4 flash is run using Q2, then people should run this one using Q½ or maybe Q¼
172.
by kzrdude 3mo ago
OpenRouter's ToS also seems to allow them to store your submitted prompts anyway, so privacy advocates would have to look elsewhere anyway, that's at least how I understand it (and it surprised me).
173.
by kzrdude 3mo ago
Google has released some very good and useful models too, Gemma, but they are not so large. But they are in useful size.
174.
by kzrdude 3mo ago
What is it useful for at slow speed?
175.
by kzrdude 3mo ago
Have defined behaviour on any input, i.e not causing security issues, memory safety issues on untrusted inputs. Lz4 has this.
176.
by kzrdude 3mo ago
ChatGPT is a household name. And OpenAi is actually not, people outside tech don't necessarily know it.
177.
by kzrdude 3mo ago
I didn't know that Amish thought so lowly of their own language, I think that's just sad. It's their own language and there's no reason to measure it against others.
178.
by kzrdude 3mo ago
Someone posted this fork, which should work in just 10 GB RAM, at least for another big model. I thought of your request and came back to spread the news :) https://github.com/ErikTromp/colibri-hy3
179.
by kzrdude 3mo ago
How many tokens/s do people feel are enough? Even 5 tok/s would be hard to use in an interesting way, not sure what the use is.
180.
by kzrdude 3mo ago
It might be impossible to run such a large model on the laptop, if I were you I'd first start experimenting with running smaller models, using llama.cpp or lmstudio, just to start understanding how it works and learn about some of the

More ›