AI has made product creation cheap, but distribution is as hard as ever. We argue that the asymmetry is not a transient feature of juvenile tooling but is structural, and four theoretical traditions converge to explain why. The argument starts with signalling. Spence (1973) models the credibility of a signal as a ratio of cost differences between high-type and low-type producers, not an absolute level of differences. AI does not zero out the ratio by zeroing out costs for all; it strips away an axis that production cost used to live on, and forces separation onto the axes that remain: track record, accumulated relationships, reputational stake. Each of these is time-indexed, and each lives in distribution, not in the product. What follows is a quality-credibility paradox. Average output quality may improve, but the credibility of any signal the product itself carries diminishes, worsening the Akerlof (1970) adverse-selection problem for buyers reached through cold, low-trust channels. This is a prediction, not a result, and it will have to be tested; it is most testable in content-heavy markets. The other three traditions show where the signalling goes. Attention economics, from Simon (1971) through Goldhaber (1997), Davenport and Beck (2001), and Heitmayer (2025), and taking seriously Bruineberg's (2025) critique of its cognitivist roots, holds that reception is gated by a budget biology fixes and supply cannot grow. Organisational ecology implies that distribution channels are constructed niches built up over time, so the liability of newness bites irrespective of the quality of the product. In speech-act theory, a trust-constituting performance is one where the speaker builds up standing; the author adds that final step as his own extension of the performativity literature, not a claim derived from it. The four converge at a point, the vouch, where someone with standing already carries out the trust-act for a newcomer and lends credibility the newcomer cannot yet generate. At that point the cost account and the speech-act account are describing one event in two languages. The bottom line is blunt: channel architecture, demand-side trust and credibility-signal construction now deserve the kind of attention founders once paid to product-market fit. Three limits, plain and simple. The framework is conceptual rather than empirical. The convergence is argued in prose rather than formalised. And the boundary of the venture class it covers is asserted and needs sharpening. The most promising next steps are formalising the cost-migration condition in a signalling model and testing the credibility-paradox prediction in a setting where AI adoption is already high.
Krishna Chagti (Sat,) studied this question.