I think local and agents is the key part of this topic. A practical approach is to define the outcome you want, test it with a small example, and then check quality, speed and reliability separately. The original point — "I wanted to start a broader creative experiments thread around Local AI...