/ writings ·

After AGI, taste is the only moat

When generation is free, the scarce inputs are taste, provenance, and intent.

By Abhishek Krishna

We spent twenty years compressing distribution. The next twenty will compress generation. What stays expensive is the bit machines can't fake: the judgment of what to make and why.

There’s a familiar two-step in the history of technology. First, a thing that was expensive becomes cheap. Then everyone wonders what’s still valuable.

Books became cheap; bookstores died and writers got Substacks. Music became cheap; labels weakened and artists got streaming. Distribution became cheap; networks ate everything and creators got an audience. Each step compressed the layer underneath until that layer was free, and the value moved up.

We are about to compress the next layer: generation.

What “free generation” actually means

In the post-AGI economy (and we are already living the early version of it), the cost of producing a competent piece of media trends to zero. A song. An essay. A photograph. A 3D character. A short film. None of these are zero today, but the rate of compression is fast enough that pricing strategies based on production scarcity are about to evaporate.

This is good news for everyone except the people whose business model assumed the opposite.

What stays expensive

When generation is free, three things become disproportionately scarce:

Taste. The judgment about what to make, not how. The ability to point at one thing in a sea of generated possibilities and say: this. This one.

Provenance. Where the work came from, who pulled it through what filter, what real human or institution stood behind it. Provenance is the only signal an audience has when they can no longer distinguish the work itself.

Intent. Why this work exists. What it’s for. Who it’s pointed at. Why now. None of these are visible in the bytes; all of them are visible in the artifact.

In a world where the bytes are free, taste, provenance, and intent are the entire P&L.

Fig. 03 · Two compressions, and the layer that won't compress expensive free 2000 2010 2020 2030 Distribution Generation Taste · provenance · intent
Two cost curves already in motion: distribution (solid, collapsed across the 2000s) and generation (dashed, collapsing now). The mint line is the layer that doesn't compress: judgment about what to make. When the cost of production goes to zero, every dollar in the system moves to whoever is supplying that flat line.

What this means for creator networks

If you take this seriously, then the architecture of a creator economy in the post-AGI era needs to invert. Today, networks pay for distribution: views, plays, downloads. Tomorrow, networks have to pay for the upstream signal: the human who spent their attention picking, framing, and committing.

That is the bet behind Create Protocol: a decentralized network whose primary economic primitive is rewarding the creative act, not the distribution event. And it’s the bet behind IPTO, which insists that licensed provenance is the floor under any of this.

You can call this idealism. I’d call it reading the tape. The cost curve is unambiguous; the only open question is who builds the rails.

A quieter version of the same point

The shorter, more personal version: if you have spent any time making something (a track, a startup, a body of writing), you already know that the expensive part was never the producing. It was the choosing.

The market is about to discover what makers have always known.