Hey guys, I'm the lead maintainer of an opensource project called StenoAI, a privacy focused AI meeting intelligence, you can find out more here if interested - https://dev.to/ruzin_saleem_2b30cf073fd8/stenoai-privacy-focused-ai-meeting-intelligence-380-github-stars-1000-downloads-4o1m . It's mainly aimed at privacy conscious users, for example, the German government uses it on Mac Studio.
Anyways, to the main point, saw this benchmark yesterday post release of qwen3.5 small models and it's incredible, the performance relative to much larger models. I was wondering if we are at an inflection point when it comes to AI models at edge: How are the big players gonna compete? A 9b parameter model is beating gpt-oss 120b!!
Top comments (3)
Awesome work with Steno, we're having a common goal regarding privacy. Whereas you do it in terms of AI, I intend to make a private blockchain, but not like Mythical, NIGHT or MIDNIGHT or even Z-Trash.
Happy to see updates on StenoAI, and will follow whenever there are some updates perhaps on Linux version. Definitely will be the one willing to test it. Personally currently I use duck.ai as my AI usage and I use it mostly for purposes of understanding the cipher structure better like for instance I was really bewildered by how AES treats it's key-schedule in comparison to DES and also as some mistake spotter and if the output makes sense I basically change the code.
I don't use discord, due to privacy concerns and connections with Persona and collecting data, so unfortunately I will not join your discord.
Rooting for you !
thank you @luftietheanonymous , will build a linux version for sure.
Github - github.com/ruzin/stenoai
Discord - discord.gg/DZ6vcQnxxu