TypeSafe AI launches Jev decision model at $0.042 per million input tokens

Positioning a model as "System One" and pricing output at zero pushes the cost of machine judgment toward the marginal cost of input alone, which is the Jevons wager made literal: if deciding gets cheap enough, software will decide far more often. The notable part is distribution — integration by OpenRouter and Cloudflare suggests the target is not end users but developers embedding decisions inside existing workflows, where a per-call cost that rounds to nothing changes architecture more than it changes budgets. Whether that substitution actually expands total judgment demand, rather than just shifting spend away from pricier models, is the open question.