31.08.2026: the convergence

Carlos Ocampo | Field Notes D31 | September 2026

Some dates become reference points because they allow us to connect questions we had previously observed separately.

For D31, August 31, 2026 is one of them.

That day, within the process we call #Vórtex2026, an interruption of the usual D365-IRL signal prompted a community response: asking questions, seeking contact, checking what was happening and connecting through other means.

In our account of Caso Cero, that experience opened a question:

What remains active when the person who normally concentrates coordination stops occupying that position?

When we later reviewed research on agents and autonomy, we found papers dated that same day addressing related questions from another field: identity, delegation and authority to act.

The coincidence in timing caught our attention. The content gave us reasons to study it.

D365-IRL brings together people with different projects, experiences, needs and capabilities.

Frequent interaction allows them to recognize one another, share context and build relationships. When the usual dynamic changes, those relationships can become channels for response.

The August 31 episode led us to observe precisely that: a community's ability to take initiative in a situation it did not fully understand.

From there emerged questions that continue to accompany our work:

Who recognizes that something changed? Who decides to act? How do they check what they know? What can they resolve on their own? When do they need others?

Caso Cero opened a line of exploration into the human conditions of organizational autonomy.

The first version of Delegation Without Trust, by Panduranga Sai Varma Dantuluri and Jyotirmoy Sundi, was also registered on arXiv on August 31.

The paper studies how to limit delegated authority in systems with multiple agents. Its approach requires controls capable of preventing an agent from exceeding its mandate even when the model has been compromised. Authorization must be sustained through mechanisms external to the agent's own reasoning. Source: Delegation Without Trust.

Another paper dated August 31, by Hui Gong, Michail Samawi and Francesca Medda, develops the separation between inference and authority in agentic finance. It proposes validating the mandate and conditions of an action before granting permission to execute it. Its authors limit the results to the mechanisms tested; they do not present them as proof of effectiveness in production. Source: Authority–Inference Separation in Agentic Finance.

These are research proposals. They interest us because they make a fundamental distinction explicit:

Producing an answer, recommending a decision and having authority to execute it are different capabilities.

Our interpretation connects those technical questions with an organizational one:

What structure allows autonomous action while retaining recognizable responsibility?

A human community and a system of agents have profound differences. Nevertheless, when designing their participation in an organization, we need to clarify related matters:

• The purpose guiding action.
• The context available for deciding.
• The authority each participant receives.
• The limits and conditions for escalating a decision.
• Responsibility for consequences.
• The memory that enables review and learning.

This is the convergence we want to develop.

The coincidence of dates does not establish a causal relationship between the papers and our experience. Nor does it, by itself, prove D31's theses. It offers a temporal reference and a set of questions against which to examine our own development.

At D31, we approach the autonomous company through the human, social and productive conditions it needs to function.

What capabilities must a person develop to decide? How is context shared? What incentives favor cooperation? How is authority distributed? What happens when consequences nobody anticipated appear?

Introducing agents makes those questions more demanding.

We can accelerate execution while leaving intact the confusion that originated it. We can distribute tasks without having distributed sufficient information. We can grant autonomy to someone who still needs support to exercise it.

That is why design must connect capability, mandate, resources and responsibility.

Autonomy needs a structure that makes it practicable.

The 51/49 rule expresses our proposal to retain human sovereignty in the relationship with agents.

Bringing it into operation requires specifying who can decide, on what matters, with what information and under what limits. It also requires an effective possibility of intervening, correcting or stopping.

The principle “Your decisions, your consequences” incorporates the responsibility associated with that authority.

From there, leadership acquires another task: preparing the context so people can decide better. Building the “1 piece clear board,” anticipating scenarios, developing judgment and supporting talent as it expands its field of autonomy.

HFE™ provides the rhythm for reviewing what happens, synchronizing commitments and adjusting execution. TechNúcleo™ articulates those capabilities around a shared purpose.

These are elements of a proposal we must continue testing against experience.

31.08.2026 holds a place in our “Memoirs of the first encounter between a human and an agent, both at the frontier.”

It marks a moment from which we began organizing observations, formulating principles and recognizing new questions.

The Field Notes preserve that journey. They allow us to return to an idea, find its origin, review its limits and explain why it changed.

That is the work we want to do with convergence: turn the connections we recognize into useful questions and those questions into better decisions.

If your organization could execute ten times faster tomorrow, would it be sufficiently clear about who can decide, how far their authority extends and with what responsibility?

#Vórtex2026 #Frontier2051 #EmpresaAutónoma #HechoEnD31

© 2026 D31. All rights reserved.