Umberto Raimondi
879
posts
@uraimo
Joined September 2008
Just FYI, there are also already IQ2_XSS quantizations available from unsloth that can be partially offloaded to make this work on <16Gb of ram, but don't expect more than a few tok/s. Not really usable.
Qwen3.8-27B is here! 🚀 It's a leap in capabilities for a laptop size model. Requires ~17GB to run locally. Model page: lmstudio.ai/qwen/qwen3.8-2…
In my opinion this is pointing in the right direction. Reading the traces/simplified reasoning can be entertaining but is it really the best thing the model can show you while it's working? Does reading a stream of consciousness really improve understanding of the current status?
You should try this if you haven't already.
Mole CLI v1.50 is here, named "Honest" 🪞 1. Clean leaves running apps alone and tells you what it left behind. 2. Uninstall finishes instead of exiting silently. 3. Failed installs name the directory at fault and the command that fixes it. 4. mo status shows purgeable space,
Nice post
Every few weeks, someone declares that coding is solved. If they mean turning a well-specified problem into runnable code, they’re increasingly right. But was writing runnable code ever the whole problem? blog.mastykarz.nl/coding-is-solv…
Wait, this is a real place? Oh man...
nice try corgicels, but i’m gonna find out what’s in that basement. double warnings never stopped the squirter
