I stopped hitting Claude's message limit by building a local AI pipeline that does the heavy lifting
…For my setup, I'm using the Gemma 4 26B —an open model from Google DeepMind—purely because it hits the practical sweet spot for the kind of Python utilities that I…