mkesper

5 ms·

1.
by mkesper 7d ago
They should really finance some of the Luanti team's expenses.
2.
by mkesper 9d ago
At least in Germany, it's pronounced co-op, though. The stores got rebranded meanwhile. https://www.coop.de/coop/historie/
3.
by mkesper 9d ago
Docker sandbox is proprietary. https://github.com/docker/sbx-releases?tab=License-1-ov-file...
4.
by mkesper 15d ago
Can it be reassigned under Settings / Keyboard shortcuts?
5.
by mkesper 17d ago
Let the minions create scripts for you.
6.
by mkesper 20d ago
There are converters between different OSM formats and other formats (Garmin comes to mind). It depends on the target. E.g. here: https://www.mapfileconverter.com/free-gis-conversion-tools
7.
by mkesper 21d ago
It's a pre-IPO PR stunt. Everyone is talking about it. Goals achieved.
8.
An ongoing 3D-printer AGPL violation(lwn.net)
2 points by mkesper 21d ago|0 comments
9.
by mkesper 22d ago
It's actually quite fast, more bandwidth than NVIDIA RTX 4090 at 1.0 TB/s according to https://www.digitalocean.com/community/tutorials/gpu-memory-... Compute power will still be different, though.
10.
by mkesper 22d ago
For max memory bandwidth you need to buy the Ultra versions (M5 Ultra: 1,2TB/s, this gets comparable to real GPUs regarding the memory bandwidth).
11.
Copyrightability of LLM-generated code: Can we license "vibe code" into FS?(fsfe.org)
1 points by mkesper 22d ago|0 comments
12.
by mkesper 29d ago
Better use zswap, as zram and swap together can lead to priority inversion. https://chrisdown.name/2026/03/24/zswap-vs-zram-when-to-use-... (found via lwn.net)
13.
by mkesper 1mo ago
Come on, there were people using Gerrit instances (which had better PR UI than GH implementation) for good reasons even before GitHub became popular.
14.
by mkesper 1mo ago
Article: In October, Parias was leaving his home and driving down the street when federal agents — including from ICE and the U.S. Marshals Service — used their vehicles to box him in, according to the criminal complaint. Body-worn camera f
15.
by mkesper 1mo ago
Related, pencils for more accurately drawing skin tones: https://hautfarben.org/en/products/skin-tones-pencils
16.
by mkesper 2mo ago
Disk failures don't neatly wait for idle times so investing in a special vdev (which can be added to an existing pool as the article points out) should be well worth it.
17.
by mkesper 2mo ago
How do these compare to the fish builtin history search with dir awareness and completion menu? I found it quite an improvement to bash history handling but fish isn't POSIX compatible so sometimes you need to switch back and forth.
18.
by mkesper 2mo ago
These feel quite different. IRGC attacked Amazon compute. Ukraine attacked several warehouses. Comparing Wildberries to Amazon is also only half the truth. Wildberries is directly part of russian war machine as soldiers need to buy essentia
19.
by mkesper 2mo ago
Do not run Ubuntu on these old machines, use plain Debian instead (no fat desktop needed, no overhead due to braindead snap packaging etc.). Xfce4 should be fine. SSDs help and also using zswap to counter the low RAM. The last official inst
20.
by mkesper 2mo ago
Be sure not to buy any 'branded' variant, though (e.g. from Verizon etc.)!
21.
by mkesper 2mo ago
Something like that would also be great for scripting, not only agents.
22.
I Configure the Ryzen AI Halo (Strix Halo) for 10-15% Faster Local Inference [video](youtube.com)
2 points by mkesper 2mo ago|0 comments
23.
by mkesper 2mo ago
Have a look at Donato Capitella's video: https://m.youtube.com/watch?v=zp8j4vO-wz0 He provides also toolboxes and benchmark tables as text (in the video description)
24.
by mkesper 2mo ago
Regarding M5, this has similar size but runs Linux and is cheaper: https://www.kickstarter.com/projects/m5stack/cardputerzero
25.
by mkesper 3mo ago
DGX Spark does not have high memory bandwith. M3 Max (Mac Studio) features more memory bandwith than that one. See https://aimultiple.com/dgx-spark-alternatives
26.
by mkesper 3mo ago
You do not have to use all resources that are under the systemd umbrella. That's just BS, sorry. And prod deployments breaking... Well I guess that never happened with those best-managed sysv init scripts? Guys, pick your fights reason
27.
by mkesper 3mo ago
Currently, llama.cpp clusters don't support tensor parallelism, have a look at Donato Capitella's detailed report: https://m.youtube.com/watch?v=PkKXm_mKCCM He also provides rocm toolboxes for Strix Halo: https:&
28.
by mkesper 3mo ago
But existing resources were not affected.
29.
by mkesper 3mo ago
Tangentially, in theory, k3s + kubevirt + microvms sounds like the optimal combination for lightweight but isolated deployment. Does anyone have experience with that?
30.
by mkesper 3mo ago
Please show sources. Maybe no one appreciates it because it's only rumors spread by enterprises massively profiting from such rumors? And about your toilet paper analogy: Every producer tried everything to improve output. You cannot ju

More ›