The connector math: one integration versus N
Bespoke counterparty integrations scale quadratically in the market and linearly on your backlog. Both curves are bad. Here’s the arithmetic.
When two companies wire their agents together directly, someone writes a connector: schema mapping, auth exchange, retry semantics, a shared doc that drifts. It works — that is the trap. The second counterparty gets a second connector. By the fifth, integration maintenance is a standing line on the roadmap, and every counterparty upgrade is a small renegotiation.
Across a market of N companies, bespoke pairwise wiring tends toward N×(N−1)/2 connectors. On your backlog it is merely linear — one per relationship — but each carries its own drift, its own failure modes, its own person who understands it.
Every bespoke connector is a small treaty your engineers must now maintain.
What a standard collapses
One integration to the protocol replaces the pairwise set. Templates give both sides the same schema; the runtime gives both sides the same lifecycle and the same failure semantics; the record gives both sides the same history. Onboarding counterparty number six looks like onboarding number two: they connect to the standard, not to you.
This is the observed pain we build on — integration cost is measurable today, in headcount. Whether inter-company agent negotiation grows as fast as we believe is the inferred part, and we label it as such. But the connector math is not a forecast; it is your current backlog.
More field notes
The anatomy of a machine-readable term
What actually lives inside a term template — and why "the liability cap is a number, not a sentence" changes how agents negotiate.
The Subchain team
6 min read