TL;DR
- 01If expert reasoning becomes a portable asset, what is the distribution model? The prompt marketplaces answered wrong and the receipts are in: weak economics, zero IP protection, trivial copyability.
- 02Registries — npm, Docker Hub, Hugging Face — answered right: free public distribution builds the network; institutions pay for private registries, governance, and assurance.
- 03Capsules are registry-shaped objects: versioned, signed (roadmap), reviewed at named tiers, with lineage and attribution that survive use.
- 04The inspection paradox that kills knowledge marketplaces — you cannot evaluate it without consuming it — is answered by review tiers and seals, not by paywalls.
The prompt marketplaces ran the experiment and failed structurally. npm, Docker Hub, and Hugging Face answered the same question correctly. Why the Capsule Registry follows the registry playbook — and why signing gates its launch.
The marketplace post-mortem
The GPT Store and the prompt marketplaces ran the experiment at scale: creator payouts that rounded to pocket change, instant copying of anything good, no provenance, no versioning, no way to know whether a listing was expert judgment or an afternoon’s improvisation. The failure was structural. A prompt is unprotectable the moment it is revealed, and an unverifiable artifact cannot command a price — the buyer cannot distinguish gold from confident filler until after paying with their own time.
What registries got right
- ▸Distribution first. npm and Docker Hub gave the public good away and became infrastructure; monetization followed adoption, as private registries and enterprise control planes.
- ▸Versioning and lineage as trust. A package with a history, maintainers, and dependents is inspectable in a way a marketplace listing never is.
- ▸Hugging Face is the closest model. Model cards, named maintainers, open weights, enterprise hubs — knowledge artifacts with provenance, monetized at the institution, not the download.
Capsules are built for this shape
A capsule already carries what a registry needs: a manifest with type and version, a review ledger with named tiers, rights and compatibility profiles, a registry-listing schema, and a deterministic digest — rendered visually as the Seal, so integrity is something you can see. Signing (Merkle root, Ed25519) is on the spec roadmap and is the gate for launch: a registry of unsigned judgment would repeat the marketplace mistake with better typography.
The sequence is deliberate: private libraries inside PRAXIS first, organizational registries second, the open registry — a curated gallery before it is an open commons — last. Expert authors join when attribution is real and the verification layer is shipped, not before. The publish-with-us waitlist is open; the economics will be registry economics.
SOURCES
