Linky|Bulga

15yo Dev | 4+ Years coding, tho code got cheap because of ai, Im also a great SWE with a lot of Computer Science knowledge.

Lowk the outlier

[Linky|Bulga @LinkyDevv]

omg, why my twitter experience so shitty

like, i have 80 followers, but no one interacts with my posts

[Linky|Bulga @LinkyDevv]

3 hours
4 vulns

We going crazy today, ngl

[Linky|Bulga @LinkyDevv]

@nalinrajput23 Because it cant, lol

[Linky|Bulga @LinkyDevv]

It doesn't matter if you are the most beautiful shade of red

If their favorite color is green

Stop changing yourself for others to love you. And instead, love your own self

[Linky|Bulga @LinkyDevv]

I just wanna bedrot all day

[NightFairy @NightFairyDream]

@thsottiaux I think can made the on/off button for the 5 hour limit, as some might hope can use freely but not limit. Still you can put 5 hour limit as default, which casual people might not know about this but for more 'expert ' user can just off the 5 hour limit, which also meet your goal

[Linky|Bulga @LinkyDevv]

@mquaser5 @thsottiaux Actually, can we just get infinite usage? Ik why there are limits

[Linky|Bulga @LinkyDevv]

@kimiku07 That said, let's re write the entire codebase in rust just 4 the love of the game

[Tibo @thsottiaux]

Hi.

It is done.

[Linky|Bulga @LinkyDevv]

@thsottiaux Reset to celebrate 5.6-cyber??

[Linky|Bulga @LinkyDevv]

@MrAhmadAwais @HaseebMir91 You probably subside a lot of api. Maybe looking for deals for a bunch of infra acess to run the models would cheap a lot the cost of most open source models.

Idk tho, just an ideia.

[Linky|Bulga @LinkyDevv]

@MrAhmadAwais @alexandr_wang @finkd @AIatMeta @CommandCodeAI That would be kinda sad tho. Because I really enjoy this model. Is the best model I've been using for so cheap

Is the 300 rpm rate general? Maybe by using mutiple api keys and distributing the load you can fix this?

[Linky|Bulga @LinkyDevv]

@imnotchalk Yeah dawg it's deepseek V4 flash, I thinks this shi is free on like, openrouter and open code

[Unsloth AI @UnslothAI]

We compared 1-bit Kimi K3 to Claude Opus 5 and GPT 5.6.

We gave 4 models the same prompt:
Create a glass aquarium whose side panel develops a visible crack and then bursts.

1-bit Kimi K3 GGUF ran locally on 4x B200s at 36 tok/s.

GitHub repo: https://t.co/aZWYAtakBP

[ST @SerenaTaN5]

sell ur remaining claude subscription token

[shivam @shivamhwp]

getting a lot of this from GPT SOL and i'm scared.

[Linky|Bulga @LinkyDevv]

@prayag_sonar They overtook fable 🥀

[Linky|Bulga @LinkyDevv]

@cyber_rekk Fuck and it was rce??? Rce ain't no joke

[Linky|Bulga @LinkyDevv]

Ai is literally a glorified calculator, its just a giant dataset with a bunch of questions and answers on the form of mathematic vertices, this dataset is made from not just data you get from the internet but also feedback and others.

And all work in a way every token weight relates with an answer. So for each token you have a weight which weights more, and that is the chosen one to be spit out back to you

The ai generated image is just an image not an algorithm doesn't involve math, but generating an image with ai involves an absurd amount of math

[Linky|Bulga @LinkyDevv]

@jullerino Dawg it's just down, did u ever checked GitHub uptime? 🥀