Dispatch
Drew Breunig: before Fable, tuning your harness was a waste because the next model paper over your problems
Breunig's line, surfaced by Simon Willison on August 23, is that improving your coding harness or context strategy used to be irrational because a new model would arrive at the same price or cheaper and erase the gap, and that Fable changed the calculus. This is the economic argument underneath the harness-versus-weights results that have been accumulating, a claim about when harness investment amortizes rather than whether harnesses matter. For a solo builder it sets a decision rule: harness work pays once model improvement per dollar decelerates.
↳ Follow the thread