Re: Patch pre-submission to Sashiko

["Theodore Tso" <tytso@mit.edu>]

On Mon, Oct 19, 2026 at 01:02:55PM -0500, johannes.goede@oss.qualcomm.com wrote:
> > It's not that hard for people to set up a local Sashiko. and then use
> > either an local LLM, or their own cloud-hosted AI tokens.  Perhaps
> > that's something we can encourage.
> 
> Actually it is that hard. Even if you spend the 2000 - 3000usd
> necessary to run somewhat large-ish models locally the results are
> still quite slow (which by itself would be ok) and they catch
> a lot less then the actual Shashiko service.

I was thinking about difficulty in setting up, but it's fair to point
out that not everyone will have access to a high-end GPU or a M5 Max
or RTX/DGX Spark setup.  On the other hand, not everyone is going to
big development/build machine. either.

So we might need to consider what a hobbyist or student might have
access to, versus a corproate-sponsored emplyoee/contractor at
different companies, etc.  The former might only have a personal "Pro"
level subscription, but some other folks might have company sponsored
token budgets which even in the post-token-maxxing era, might be a bit
more generous than a subscription plan.

As someone else has pointed out, students / hobbyists just use
OpenCode with the review prompts with one of their free models, such
as "Big Pickle".  That or a local quen3-coder-next (my current local
LLM, when I'm not using my employer's cloud hosted models) might not
be as good as Sashiko, but you can find a lot of the obvious issues
first, before you send it to the public list.

	       	      	       	     	 - Ted