Quote from OpenAI on the livestream.
'Already, Sol has been transforming our research program. As one example, GPT-5.6-Sol autonomously post-trained GPT-5.6-Luna.'
One of the best use cases I have found this is generating CUDA kernels - either directly based on a problem statement or looking through Torch profiler traces and identifying opportunities for optimizations.
The model is extremely good at going on in a loop where it collects a torch profiler or Nsys trace, reads through them, identifies the shapes and sizes the different ops go through (and their number of instances as such), reads through the Nvidia documentation/Programming Guide for relevant parts and then writes a kernel that is performant and very clean - it can continue to iterate based on the targets.
I have some harder tasks like FA4 and such in the pipeline which I want to get to.
This "loop" automation is nuts inside of Codex.
"/goal go over every single feature in this app create a user story with expected behaviour based on the code keep a single canonical spreadsheet tracking the features status
- when done switch loop to testing every user story and documenting all errors
- when done fix every logistical error or ux error
- test every user behaviour again post fix"
Shoutout to @MatthewBerman for the heads up.
Hundreds of user stories being worked through like it's nothing.