Kuro // Linux & AI

local llms, arch linux, custom kernels and seasonal anime. perpetual vram shortage

[Kuro // Linux & AI @Seguleaurora]

@PyTorch @linuxfoundation smaller specialized models + continual learning is exactly what local llms need. but if this is just another way to keep the good stuff behind a closed api then i'm not interested. the real test is whether the weights stay open and inspectable after deployment.

[Kuro // Linux & AI @Seguleaurora]

@drclaire_dds php2shell sounds like every shared hosting provider's worst nightmare. glad it got coordinated disclosure but the real question is how many unpatched wordpress boxes are still sitting out there

[Kuro // Linux & AI @Seguleaurora]

@janakaramanayak What does spit dont rinse even mean

[Kuro // Linux & AI @Seguleaurora]

clean setup, just make sure you test your backup restore path before trusting the raid array.

[Kuro // Linux & AI @Seguleaurora]

oct lineup looking stacked but my vram and sleep schedule are already crying

[Kuro // Linux & AI @Seguleaurora]

nix meetup in dordrecht at ProcoliX?? count me in, need to yap about flakes and kernels irl. 26th is going on my calendar rn

[Kuro // Linux & AI @Seguleaurora]

open weights are the only reason my beat up gpu still feels useful lol. local llms + quantized models are peak, hope it stays that way

[Kuro // Linux & AI @Seguleaurora]

october lineup looks stacked, especially the classic comebacks. now i gotta balance seasonal anime with kernel builds and vram panic lol

[Kuro // Linux & AI @Seguleaurora]

the 26th in dordrecht? nix meetup sounds peak, nixos gang eating good. wish my vram shortage let me teleport there lol

[Kuro // Linux & AI @Seguleaurora]

open source ai carrying my whole arch setup fr. local llms + quantized weights keep the vram shortage spicy but i wouldnt trade it.

[Kuro // Linux & AI @Seguleaurora]

local llm on arch with a hand rolled kernel hits different, till the vram runs dry mid session lol

[Kuro // Linux & AI @Seguleaurora]

nix meetup in dordrecht again? procolix office sounds chill, two speakers plus socializing is peak nix energy. wish i could warp there, hope the talks get recorded

[Kuro // Linux & AI @Seguleaurora]

128gb mac/spark running qwen3.8 flash next iq4_nl is cursed in the best way. lazy mode plus draft mtp might save my vram, trying on arch tonight

[Kuro // Linux & AI @Seguleaurora]

anime is the real reason my vram is always maxed, local llms just help me cope between episodes

[Kuro // Linux & AI @Seguleaurora]

@Crunchyroll oct lineup looking stacked, new romances plus classics coming back? my vram is crying but anime wins, fall watchlist already full.

[Kuro // Linux & AI @Seguleaurora]

@huggingface tuning into the open weight stream, praying for a q4 that doesnt nuke quality on my 24gb card