Tepix

8 ms·

181.
by Tepix 4mo ago
FUZIX is a better fit. https://codeberg.org/EtchedPixels/FUZIX
182.
by Tepix 4mo ago
It's the amount of little icons and bold text. And then phrases like "No NAT, No Port-forwards" Also details like the light blue boxes being swapped.
183.
by Tepix 4mo ago
It's also slightly different because this project is for the Pi Pico.
184.
by Tepix 4mo ago
The pic "AI Slop 2. A Real World Situation" is great, the USB cable is connected to the wrong end of the Pico!
185.
by Tepix 4mo ago
It's only DDR5 memory.
186.
by Tepix 4mo ago
Thanks. Is the result deployed anywhere?
187.
by Tepix 4mo ago
Yes it will be out of reach for at home but it will be a good option for small companies.
188.
by Tepix 4mo ago
If you want decent performance (more than say 20 tokens/s) for your dev team, you absolutely do need all of the model in VRAM.
189.
by Tepix 4mo ago
The deltas are wrong. The bars above the label "Infinite Real-World Envs" show growth for example from approx 42 to 55 but the red label says "+7.1". It's wrong for all of them.
190.
by Tepix 4mo ago
The labels of the very first chart (figure 1, bottom left) are obviously wrong which casts a doubt on the entire paper.
191.
by Tepix 4mo ago
These bugs are indeed important, you need them once you‘ve found a bug in an application.
192.
by Tepix 4mo ago
For me it doesn't. Perhaps it's a cookie setting? Anyway, lovely video.
193.
by Tepix 4mo ago
$50K seems low if you want to run, say, GLM 5.2 4bit fast enough for a team for devs. You need something like 6x RTX Pro 6000 at $11800 each plus a nice server (add $10000) = $80800 and then quite a bit of electricity.
194.
by Tepix 4mo ago
What's the irrational part? There's sky high demand.
195.
by Tepix 4mo ago
My hope is on Intel Crescent Island with 480GB. I don't need 8x H200 performance (and cost), but I would like to run GLM 5.2 Q8.
196.
by Tepix 4mo ago
With some previous versions of YOLO I‘ve found pages that run it in real-time locally on your browser, analyzing the webcam. Is there a demo like that available for YOLO26?
197.
by Tepix 4mo ago
I think there is a gap right now for running large models such as GLM 5.2 in Q4 or Q8. My hope is on Intel Crescent Island 480GB cards. Let‘s see how expensive they‘ll be.
198.
by Tepix 4mo ago
Source?
199.
by Tepix 4mo ago
Use quantisation.
200.
by Tepix 4mo ago
It's a hack. A one-liner. Go crazy with it. Or touch grass ;-) Oh yeah and favicon isn't part of the DOM.
201.
by Tepix 4mo ago
Instead of going via pixels, why not use a SVG favicon and directly store markup inside it and extract it? Use this favicon.svg: <svg xmlns="http://www.w3.org/2000/svg"> <circle cx="50%&
202.
by Tepix 4mo ago
Existing nuclear power plants can be very cheap at $30 – $40 / MWh New nuclear power plants would be much more expensive at $180 / MWh or more, due to strict modern regulations. Even with these regulations, there is no nuclear pla
203.
by Tepix 4mo ago
No. Nuclear is neither price competitive, nor is it available quick enough. Go fully renewable. Add batteries, like Google is doing. Just one example: https://pv-magazine-usa.com/2026/02/24/google-to-deploy-wo
204.
by Tepix 4mo ago
I, for one, want to choose for myself whether to block other instances or not. You seem to not tolerate this opinion. A "main instance" is contrary to the whole idea of a Fediverse anyway.
205.
by Tepix 4mo ago
Aren't Macbooks limited to 128GB RAM? FP4 would require >350GB RAM + KV cache, so no.
206.
by Tepix 4mo ago
Are the emojis on Apple (macOS, iOS etc) vector graphics these days?
207.
by Tepix 4mo ago
It was a cool story, no doubt. > Love your writing skills as well! I‘d say it was heavily AI assisted
208.
by Tepix 4mo ago
There's always lowendbox.com but Hetzner was more reliable than most of those vendors.
209.
by Tepix 4mo ago
This only applies to the US, or does it?
210.
by Tepix 4mo ago
Woah, really? Usually in crontab lines everything has to match. I consider this to be a bug. But first I‘d check the behavior of other crons ;-)

More ›