Creation Got Cheap. Discovery Didn't.

The cost of creating things is collapsing. Suno lets someone with no music production training generate a full track. Descript turns rough footage into a polished video without a professional editor. Midjourney, Runway, dozens of writing tools. A person with taste and a laptop can produce what used to require a team and a budget.

This is real and it matters. But it solves the wrong bottleneck.

In music streaming, roughly 90% of catalog is never heard. This isn’t a guess. Platforms like KKBOX, which operates across five Asian markets with over 10 million active users, see this directly in the data. The library is massive, the listener’s attention is finite, and the infrastructure for connecting the two hasn’t kept pace with the supply side. The problem is structural, not editorial. You can’t curate your way out of a catalog that grows faster than any human team can review.

AI creative tools are about to do this to every medium. More music. More video. More writing. More of everything, produced faster, at lower cost, by more people. The supply curve goes vertical. But discovery infrastructure has not had the same leap.

The result is predictable: noise. Not in the pejorative sense. A lot of what gets created will be good. Some of it will be excellent. But “good content exists” has never been the problem. The problem has always been “the right person encounters it at the right time.” That problem just got much harder.

Search doesn’t solve this. Search requires you to know what you’re looking for. Recommendation algorithms help, but most optimize for engagement, not quality or fit. YouTube’s algorithm surfaces what keeps you watching. Spotify’s Discover Weekly optimizes for streams. These are reasonable proxies, but they reward retention, not resonance. Editorial curation is labor-intensive and doesn’t scale. Each approach handles a piece. None of them handle what’s coming.

The interesting question for the next wave of media companies isn’t “how do we help people create more?” It’s “how do we help the right work find the right audience?” That’s a fundamentally different product problem. It requires understanding context, timing, trust, and taste in ways that current recommendation systems don’t attempt. The work I’ve seen in streaming personalization, designing feeds that balance familiarity and surprise based on the listener’s actual moment, is a small version of this. The full version is still unbuilt.

The companies that figure it out won’t look like creation tools. They’ll look like something between a curator, a distribution platform, and a matching engine. The value they create won’t come from what they help people make. It’ll come from what they help people find.