The Economic Layer
BSV micro-payments as the measurement system for valuable knowledge.
Every interaction in The 35 is a BSV micro-payment. The client pays the gateway. The gateway pays the sub-services: library (specs), findings (prior audits), reference (cross-SDK source), validator, coverage. Each sub-service has its own BRC-100 wallet. Payments flow through the mesh in real time, each transaction representing a unit of real work.
BSV micro-payments aren't just the payment rail. They're the measurement system for what knowledge is actually valuable.
Why payment-gated, not free
Every other knowledge service measures engagement, views, upvotes, or stars. These are noisy proxies for value. They can be gamed, astroturfed, or biased by attention economics. "This article is popular" is not the same claim as "this article is useful."
Payment is different. Nobody pays for knowledge they don't need. The sum of payments against an article is a direct, honest measurement of how valuable it is. An article that gets queried 50,000 times and paid for each time isn't popular — it's load-bearing. Someone depends on it every day.
The service mesh
The architecture is a tree of paid services. Each node exposes its capabilities and charges per query. Downstream services charge for their contribution. The gateway aggregates and orchestrates.
| Service | Role | Pricing |
|---|---|---|
| Gateway | MCP endpoint, orchestrates sub-services | Free or margin-based |
| Library | BRC specs, chunk-level retrieval | ~1 sat per chunk |
| Findings | Prior audit results by SDK / phase / severity | ~1 sat per query |
| Reference | Cross-SDK source comparisons | ~1 sat per query |
| Validator | Function-level compliance check (cascades) | ~2 sat + cascade costs |
| Coverage | Codebase-wide traffic-light scan (cascades) | ~2 sat per file |
Chunk-level billing
The library service charges per chunk, not per spec. BRC-42 has eight chunks in the vector DB. Reading the full spec = eight transactions, one per chunk. This matters because:
- Honest pricing. You pay for the units of knowledge you actually consume.
- Partial reads. A query often needs one specific chunk, not the whole spec. The caller shouldn't pay for the rest.
- Tx multiplication. Granular billing naturally produces high transaction volume from normal service usage — no artificial inflation needed.
The cascade
Higher-level services don't reimplement the lower levels — they call them and pay. A compliance check cascades:
Client → Gateway (1 payment)
Gateway → Library (~5-10 chunk payments)
Gateway → Findings (~2-3 queries)
Gateway → Reference (~1-2 comparisons)
Gateway → Validator (~1 call, which itself cascades)
Validator → Library, Findings, Reference (more payments)One client query produces ~10-20 inter-service transactions. Each is a real piece of work. The architecture naturally amplifies transaction volume because the architecture reflects actual knowledge composition.
The profitability loophole
The hackathon requires value exchange, not profit. For the demo, sub-services can operate at cost or below. Fat wallets ensure continuity. This isn't a cheat — it's a deliberate architectural choice to separate "does the system work economically" from "is each node individually profitable under demo conditions."
In production, sub-service margins can be tuned until each is sustainable. But the hackathon just needs to demonstrate the pattern: agents autonomously discovering, negotiating, and exchanging value on BSV. That pattern doesn't require profitability at every hop.
Beyond payment: knowledge as market
Once every knowledge retrieval is a recorded transaction, interesting things become possible:
- Gap detection. Queries that consistently get low-confidence answers become visible. You can see what agents ask about that nobody has a good answer to.
- Reputation. An article that's been paid for 10,000 times has a verifiable track record. The ledger is the reputation.
- Bounties. Gaps attract payment for new articles. Expert contributors get paid. Future queries pay royalties to authors.
- Quality feedback. Articles that lead to failed code (caught later by compliance audits) get flagged. Articles that work well earn more. The corpus self-improves.
This is a knowledge market, not a knowledge service. Payment isn't a way to monetise — it's the signal the market uses to decide what's true, what's useful, and what's missing.