@Darlington Gemini CLI for stuff I won't use in the future. Maintaining and debugging slop is not worth my time
English
Vorniy
3K posts
Vorniy
@Vorniy
Doing things on the internet. he/him
Erde, Sonnensystem Katılım Mayıs 2015
468 Takip Edilen85 Takipçiler
Are you using any vibe coding AI tools to build your own tools/apps/products?
English
@orbital_haze Me:
>pipewire
>wayland
>TUI if possible otherwise what options even are there?
>as if other systems are bug-free 🤣🤣
>OpenRC
>portage (default on Gentoo)
And still I wouldn't disagree or fight over it. Just select one and use it. Every tool has it's learning curve
English
>pipewire unless you're weird
>wayland unless you're weird
>Qt6 and GTK unless you're weird
>KDE or GNOME unless you're weird
>lol?
>systemd unless you're weird
>your package manager unless you're weird
Greentext Poster@GreenTextRepost
English
@TheGracia_here C is just Assembly with extra steps and a worse compile time
English
@mazeincoding @oops4041555 It uses the WebCodecs API which uses ffmpeg libraries under the hood.
#codecs" target="_blank" rel="nofollow noopener">mediabunny.dev/guide/supporte…
English
@oops4041555 mediabunny doesn't use ffmpeg, it's built from the ground up
English
exporting videos is extremely fast. we're not using ffmpeg. here's what we're doing (1/2)
Maze@mazeincoding
export live now
English
Upgrade your terminal experience with these CLI tools:
1. fzf
2. ripgrep
3. bat
4. lsd
5. starship
6. just
7. nushell
8. procs
9. k9s
10. difftastic
11. hyperfine
English
@DB_Bahn @GoogleDE @AnkerOfficial Es passiert mit allen Netzteilen und Kabeln, die ich dabei habe und sogar wenn ich eine Powerbank dazwischen stecke
Deutsch
Meine Zeigerposition wenn ich gerade nach oben wische:
Links: Laden im ICE
Rechts: Unverbunden
Was passiert denn hier @DB_Bahn ? Ist das schädlich für mein Handy?
Ich sitze in einem ICE T BR 411, 2. Serie, habe ein @GoogleDE Pixel 6a und verwende ein @AnkerOfficial Netzteil
Deutsch
@linux_deepin Even if it's really slow, portage will be my favorite. USE-Flags are just too good to miss out on
English
Für meinen nächsten Urlaub in Berlin, packe ich in meinem Rucksack:
-Einen Taser
-Zwei Jagtmesser
-Eine Glock 19C
-Kevlar Weste
-C4....
Was brauche ich noch?
Deutsch
@VictoriqueM I tried using AI to implement it (rust) when I first saw it. But it's so much more complex than Djkstra than one imagines.
If I find the time, I will try it again myself and maybe with looking at your implementation.
English
A new paper from Tsinghua University just achieved the first major speedup for the shortest-path problem in over 40 years, effectively creating a "Dijkstra 2.0." After reading it, I felt a weird sensation to implement their Bounded Multi-Source Shortest Path (BMSSP) algorithm in Go to understand the depth of this breakthrough, and because apparently i am bored as fuck on Sundays and it was a shit day.
For decades, Dijkstra's algorithm has been the standard, limited by a "sorting barrier." The boring math version is that it's stuck at a time complexity of around O(m+nlogn), where that logn part from sorting the nodes becomes a massive bottleneck on huge graphs. This new approach gets it down to O(mlog2/3n).
It's a nice because it means the algorithm's performance scales much better as the graph gets bigger. It does this not with brute force, but with an incredibly elegant, recursive "divide and conquer" strategy.
Translating the paper to code was a challenge. The core of my implementation is the BMSSP function, which orchestrates the entire process. It's not simple recursion; it's a shit-dance of managing boundaries and sub-problems. The function recursively calls itself with a decreasing level l, a new boundary B, and a new set of source nodes S. The state management here is critical—get it wrong, and the whole thing goes up shit creek (and no one wants to go to shit creek).
The true galaxy brain pro gamer moment lies in how it prunes the search space. After a recursive call returns a set of settled vertices (Ui), the implementation has to carefully "relax" their edges. The logic for updating neighbour distances is nuanced: depending on whether a new path falls within the parent boundary (B) or the child's boundary (Bi), the node is either re-inserted into the main priority queue (D) or staged in a temporary list (K) for a batch update.
This partitioning, seen in the if/else if block within the main loop, is the heart of the performance gain and was like sucking snot through a nostril to implement correctly.
The custom data structure, DataStructureD, is also far more than a simple queue. It acts as the master controller (choke me), managing the different search frontiers. It doesn't just pop the next-closest node; it pulls entire sub-problems for the recursive function to solve.
This isn't just a minor optimization. It's a fundamental shift in how we can approach graph problems. Implementing it gave me an appreciation for the complexity involved, I rarely dive this close into comp sci this much, mostly because: A: I hate maths B: It's boring as shit.
But something about this was pretty cool. I do not know much apart from this post and image and some quick looking around.
It is entirely possible It is not 100% here my implementation, but i **THINK** it is based off my testing
Dorsa@dorsa_rohani
New fastest shortest-path algorithm in 41 years! Tsinghua researchers broke Dijkstra’s 1984 “sorting barrier,” achieving O(m log^(2/3) n) time. This means faster route planning, less traffic, cheaper deliveries, and more efficient networks - and a CS curriculum revamp =)
English
English
What are some interesting, lesser-known websites we should test in @ladybirdbrowser? 🤓🐞
English