Blockchain proves a record, not legitimacy

Carlos Ocampo · Field Notes D31 · September 2026

An operation was recorded. We can consult its trail, identify a transaction and verify certain data.

What conclusions does that evidence support?

Was the action authorized? Did whoever authorized it understand its consequences? Was the commitment fulfilled? Were the source data correct?

Each question requires different evidence.

The paper Authority–Inference Separation in Agentic Finance, by Hui Gong, Michail Samawi and Francesca Medda, examines this separation. Its authors describe how blockchain can enforce programmed conditions and provide verifiable evidence of a transaction, while the institutional mandate, delivery of a service and responsibility require additional information and judgment. It is a research proposal with a bounded evaluation; its results do not demonstrate general effectiveness in production. Consult the study.

Our interpretation at D31: each piece of evidence must be accompanied by an explanation of what it demonstrates and what it leaves unresolved.

Consider an agent authorized to contract a service. The payment is recorded, but the provider delivers an incomplete result.

The payment record helps reconstruct part of the operation. To evaluate fulfillment, we also need to know what was requested, what conditions were accepted, what was received and who reviewed the delivery.

The same discipline applies to our field documentation. A screenshot preserves a fragment. A recording lets us review what was captured. A log organizes events from a perspective. By relating them, we can build a more complete explanation while keeping the gaps visible.

If part of the process went unobserved, preserving the rest does not eliminate that absence.

That is why we propose a case file that connects intention, authorization, action and outcome. One that distinguishes what was planned from what was executed. One that preserves versions, corrections and objections from the people involved.

In TechNúcleo™, we propose that this memory of decisions form part of organizational design. Its usefulness must be tested through a practical question: can someone who did not participate in the operation reconstruct what happened and recognize what we still do not know?

The 51/49 rule also runs through this discussion. The record must allow people to examine how authority was exercised, question a decision and request its correction.

For a company introducing agents, good documentation means being able to explain the full path: where an instruction originated, under what conditions it was executed and what consequences it produced.

Trust requires knowing what can be verified and what remains open.

If someone questions a decision by your organization tomorrow, does your evidence allow you to reconstruct it, or only show that an operation was recorded?

© 2026 D31. All rights reserved.