We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! ๐งต
Blog: https://t.co/bOWCcFhCVq
no matter what agent you're using, your voice is the best way to talk to them
and hex is the best way to use your voice:
- fast
- runs locally
- works on mac, windows, linux
stable release now available: https://t.co/Eb78egcyoP
Excited to present tomorrow at #CUI2025!
I'll be sharing our work on CommCoach, a conversational AI role-play system to support managers practicing tough workplace conversations.
๐ Paper: https://t.co/xReyn3ftfG
๐๏ธ Talk info: https://t.co/ZZTr9vIXlS
1/3
We found managers valued adaptive, low-risk simulations for learning, but also raised key concerns:
โข AI consistency vs adaptability
โข Bias in role-play personas
โข Structured goals vs open-ended convos
Designing AI tools for communication coaching is nuanced and timely.
2/3
Iโll share insights from our early work with CommCoach, a prototype AI role-play tool designed to help managers practice workplace communication. It explores the potential of adaptive, low-risk training and the tensions that come with AI-driven feedback. Hope to see you there!
Excited to be speaking at Virginia Tech's Annual Professional Development Conference on May 8! Iโll be sharing early work on using AI to support workplace communication training, especially for supervisors.