pHequals7

6 ms·

1. 1.▲
   by pHequals7 5mo ago

   Currently the on device models such as Parakeet and Whisper are great for English, faster than cloud hosted models a little less accurate - if you switch on the post processing, the ASR output goes through a fine tuned Qwen 3.5 model that i
2. 2.▲
   by pHequals7 5mo ago

   let me know if you face any issues - and always looking for more collaborators!
3. 3.▲
   by pHequals7 5mo ago

   Github: https://github.com/pHequals7/muesli Looking to add on device CUA and support more models (MSFT Vibevoice, IBM Granite etc)
4. 4.▲
   Show HN: Muesli – If Granola and Wisprflow had an open source on device baby(freedspeech.xyz)

   15 points by pHequals7 5mo ago|10 comments
5. 5.▲
   FrontierSWE – Benchmark for long horizon coding tasks(github.com)

   1 points by pHequals7 5mo ago|1 comments
6. 6.▲
   by pHequals7 5mo ago

   Interesting work from Proximal - love the focus on out of distribution tasks like git to zig..
7. 7.▲
   by pHequals7 5mo ago

   thankfully with code and coding agents - the tacit/tribal knowledge always lives via the codebase itself unlike atoms based manufacturing processes..
8. 8.▲
   by pHequals7 5mo ago

   lol pseudoscience as a service ha!
9. 9.▲
   by pHequals7 5mo ago

   unfortunately true :(
10. 10.▲
    Aspartame is not that bad? (2022)(dynomight.net)

    132 points by pHequals7 5mo ago|252 comments
11. 11.▲
    by pHequals7 5mo ago

    we are in the times of irrational exuberance - rationality will set in soon!
12. 12.▲
    Convert every software into an RL environment(cmu-l3.github.io)

    2 points by pHequals7 5mo ago|0 comments