Against Vibes: When is a Generative Model Useful
Summary
This blog post critiques the hype around generative models and proposes a pragmatic framework to judge usefulness based on encoding cost, verification cost, and whether the task emphasizes the artifact or the process. It argues that usefulness often falls with increasing task complexity, and that genuine usefulness requires domain knowledge and engineering rigor rather than vibes or marketing claims. The author encourages careful evaluation and transparent criteria to determine when generative models are truly helpful.