Running local models is good now

with the most recent releases from Google in the Gemma 4 family, I've finally been able to do agentic coding locally and have loops work at about ~75% the accuracy/speed of frontier models. these kinds of tasks, even as simple as they are, used to be impossible for local models as recently as 6 months ago.

pragmatic take on local model viability. vickiboykis documents the shift from toy to usable—not hype, just “do i have to double-check against an API?” as the metric. the piece avoids breathless claims but makes a real observation: gemma-4-12b-qat handles agentic loops well enough for actual work (refactoring, linting, test generation). includes concrete setup details (pi + lm studio in docker) and architectural tradeoffs worth considering. refreshing absence of “local models will replace everything” posturing; instead, focuses on the narrow but genuine expansion of what’s now feasible on consumer hardware.