Vicki Boykisnew post: how I develop recently using local models. the tooling is now good enough to do agentic workflows and everyone should give them a try!
https://vickiboykis.com/2026/06/15/running-local-models-is-good-now/
Dave Wilburn 
IMO, getting about 75% the performance of a frontier model from a locally run open weight model is going to absolutely destroy the business model of the big AI firms (including, ironically, Google itself). Companies looking to tighten their belts while big AI firm token costs skyrocket will be strongly incentivized to make it work.
Plus, a local model won't leak the crap out of your intellectual property, PII, or other confidential data to your Big AI service provider that might not-so-secretly see you ultimately as a competitor rather than a customer.