Legibility of effort in the LLM age

Jason Kottke points to a sharp little essay, “Legibility of Effort,” on something we’ve been circling as we build with AI: software and writing are losing their legibility of effort — the ability to tell at a glance whether something took a human meaningful work.

That signal used to be load-bearing. A dense, well-considered interface implied care; a thorough piece of writing implied thought. When a model can produce either in seconds, the glance-level heuristic breaks — and we haven’t replaced it with anything.

We think this matters for how tools present themselves. If effort is no longer legible in the artifact, trust has to come from somewhere else: provenance, a track record, the parts a person still visibly owns and stands behind. It’s an open question, and a good one to sit with as the ground shifts under it. Worth the two minutes.