We’ve worked closely with multiple AI teams, and this data lines up with what we’re seeing on the ground. Revenue doesn’t just grow faster in AI, the spread between average and exceptional execution is massive. A few observations from working inside these companies: > The

its sad that linux doesnt natively support cpu tied fan curves for both cpu and gpu for laptops that have shared cooling, BUT its cool that you can make your own daemon to do so

gmgami

we all start somewhere, simple gruvbox themed linux rice #linux #ricing

if you're using linux, why? if you're not using linux, why?

all you need to get started: - a laptop - claude code - a dream

The new Neobank Positioning Skill A 9-point output, territory map, and white space analysis built from real campaigns. Built for neobank teams who ship great products but struggle to articulate why they're different. npx skills add growgami/neobank-positioning-engine-skill

small models don't need to be smarter. they need better architecture around them 🧠 • DSPy-compiled prompts on a 770M model match GPT-3.5 • MoA with ≤9B models beats standalone GPT-4 • RouteLLM cuts API costs 85% while keeping 95% quality the gap is closing🔥

pewdiepie is awesome but finetuning a model has been around for awhile 👀still awesome tho

decentralized ai

x.com/i/article/2031…

claude is goated fr

oss is thriving with the advent of ai

a harness is greater than the agent

trying out niri and i have to say, i like it more than hyprland for my workflows

linux + ai makes you unstoppable, unless a system update breaks something

migrated from hyprland to niri on my nvidia laptop and mass suffering aside, the scrollable tiling workflow is so worth it. cachyos + wayland + nvidia is not for the faint of heart but once you get past the vram quirks and config rabbit holes, it just feels right.

mythos is crazy

bruh

i only post for testing purposes, y'all post for fun, we are not the same