Re: [MAINTAINERS SUMMIT] The place of AI code review in the Linux Kernel process

On Wed, 15 Jul 2026 21:21:03 +0000
Roman Gushchin <[email protected]> wrote:

> >> Many engineers have asked for some sort of interactive mode where they
> >> can ask additional questions or follow up on the initial feedback from
> >> Sashiko. I plan to add this to the local review mode, but for the
> >> central public instance, it is problematic from both security and token
> >> cost perspectives.   
> >
> > Could there some way to download the entire context from all the passes
> > to a local environment and have a local llm chew on it to answer
> > questions?  
> 
> Yeah, it's an option I though about. Like a download button which just
> dumps all relevant context as an .md file, for example.

The better would be to produce, instead, a trained dataset - or - even
better - a patch-review kernel-trained model with a couple of different
sizes (and kv compression), in GGUF format, allowing people to use it
directly, e.g. something that could work like this:

	ollama run hf.co/sashiko/sashiko-<some_model>:8b
	ollama run hf.co/sashiko/sashiko-<some_model>:36b-a4b


Thanks,
Mauro