News for hackers. Stuff that matters.

skye0110

skye0110 on Hacker News

Ask HN: Are third party gguf/mmproj safe on Llama in production env

Posted by skye0110 on Tuesday August 11, 2026 @01:41PM
from the ask-hn dept.

So i had been building a side project screenmind ,sort of local ai desktop assistant, and hit a confusing multimodel failure...texts worked,but vision and image failed no crash just returning <unsused49>. i narrowed it down to unsloths gemma 4 gguf +mmproj broke on llama(atleast thats my hypothesis , feel free to correct me) b10244, and the fix was to just switch to ggml.So is pinning...

( Read More... | no comments | 2 points | on HN )
Re:Show HN: Local privacy-first Microsoft Recall alternative with Gemma 4 on Saturday July 04, 2026 @03:58AM (#48783546)
thats the nice part though-not everything has to be a moat , anyone can build what they think , no idea remains a idea thats a win.

Show HN: Local privacy-first Microsoft Recall alternative with Gemma 4

Posted by skye0110 on Friday July 03, 2026 @11:30PM
from the show-hn dept.

https://github.com/ayushh0110/ScreenMind/blob/main/README.md

hi author here, Screenmind is privacy first Microsoft recall alternative . It runs on gemma 4 which is one of the fewer models supporting vision audio and reasoning all 3, so your data never leaves you machine.With screenmind you can keep a track of your timeline , how much time you spent on what..search any screenshot with any text on it.. and the coolest thing, you can chat with your screen...

( Read More... | 2 comments | on HN )
Re:Show HN: I run a vision model on every screenshot, locally, on a 4G... on Monday June 15, 2026 @02:18PM (#48545105)
agreed, the backlash was the cloud + always-on , not the idea itself. thats basically why i only bothered building the local version.
Re:Show HN: I run a vision model on every screenshot, locally, on a 4G... on Monday June 15, 2026 @02:17PM (#48545094)
thats the point - local, opt-in, nothing leaves the machine. rewind was great but the cloud part scared people off. this runs the whole analysis on-device so the search/chat works without anything being uploaded and it works more as a background service.