Higher Self

I believe communities should own and govern their own intelligence.

It’s an honour to collaborate with UNSAAC and its LAAD lab on El Hilo que Persiste, a programme supporting communities to own their communication infrastructure: what it’s for, who can use it, what it records and when it stops.

Read the field report, in Spanish hilopersiste.pe
The field team at the entrance sign to the Ausangate private conservation area, in the Cusco region.

The first field test.

In July 2026, at about 4,300 metres in the Cusco region, we simulated a visitor on a trail sending a call for help. It crossed the hillside by long-range radio, with no internet and no mobile network, and reached the village plaza.

The bell rang. That was all the prototype had to show: that a sound can cross a hillside and reach someone.

What we built
Something deliberately small: long-range radios and a weatherproof box with two key locks, running a local Wi-Fi network, an app to send and answer a call for help, an append-only case log and a bell. Nothing depends on the internet, a mobile network or a server someone else owns.
Who
Five UNSAAC students in engineering, informatics and anthropology came as independent auditors, to check whether what we said was true. Orietta Trebejo coordinated the fieldwork and the relationship with the community. David Bentley built the prototype.
Agreement
The community’s assembly reviewed the project and agreed to go ahead.

What the test taught us.

We went to learn the terrain, not to prove the system worked. Most of what we learned was about people, not radio.

Evidence
The message arrived and the bell rang, at an estimated 300 to 400 metres.
Not claimed
Coverage. The distance was judged by eye, and the link was point to point, not yet a network.
  • The bell has to wake someone asleep at night, in the cold. Ours wasn’t loud enough.
  • A responder who left the app couldn’t get back in while the alarm kept ringing. Locking out someone who’s already trying to help is the failure that matters most.
  • The node’s clock ran ten days behind. A hash fixes content, not time.

Every request the community made was about being able to act, not just to read: see which cases are being handled, tell cases apart, and say “I’m handling this one”.

“Today we install nothing.”

That’s how our letter of intent to the community begins. It was written to be read aloud, so the assembly could decide with the whole picture in front of it.

  • If the community goes ahead, the system is theirs: the name, the data, the keys and the decisions, to use, switch off, change or withdraw without asking anyone.
  • The programme funds and maintains the full system for the first year, and David commits to three years of training and support, so the community can run and repair it themselves.
  • In return we ask for ayni, Andean reciprocity: that the community governs the system, looks after it, and says honestly when it fails.

What comes next.

With the assembly’s approval, we’ll walk the area with the community to map where nodes, repeaters and trail beacons should go, then bring the production design back to the assembly before anything is built.

Not claimed
That it’s an emergency service. It’s a prototype in development, and nobody should rely on it as one.

The papers behind it.

Before anything went up the mountain, the rules were written, published and opened to scrutiny: six documents between December 2025 and March 2026, moving deliberately from ethics to technical controls.

The ethical base, December 2025

Framework Protocol for decentralised infrastructure and sovereign governance in Indigenous territories

Free, prior, informed and continuing consent is mandatory, and “no false security” is an absolute limit.

Read it, in Spanish
The governance standard, January 2026

Reciprocal Intelligence (v0.1)

Community authority as an enforceable technical requirement, not a consultative preference.

Read it, in Spanish
The technical base, January 2026

IR v1.0: sovereign infrastructure and distributed intelligence

How offline-first hardware must be built and secured. A full specification and a public summary.

Read the public summary, in Spanish
The evidence protocol, February 2026

Sovereign Khipu Protocol (PKS v1.1)

Tamper-evident records made offline, with append-only, hash-linked logs and physical kill switches. If a piece of proof is missing, the answer is “not proven”.

Read it, in Spanish
The custody model, March 2026

PNT, Par Nodo Tinkuy

Two separately held nodes that create a shared record of an event offline, reconciled in person rather than through a central server. Its rule: if it can’t run without taking the community’s keys, its ability to switch off, or its legitimacy, it must not run.

Read it, in Spanish
The field report, August 2026

Between the specification and the deployment

The first field test in full: what we prepared, what we built, what the test found, and what changes next.

Read it, in Spanish
Not claimed
Independent verification or peer review. None of these documents has yet been checked independently against the prototype, and no institution has validated the system. They’re working papers and draft standards, published openly.
All publications on ORCID The papers are in Spanish, the language of the work. The English summaries here are ours.

What would you like to be easier?

Tell me one thing that’s slower or shakier than it should be. A sentence or two is enough. I reply within one working day, on New York time, and I won’t chase you. Please leave out passwords and customer records.

david@higherself.ai WhatsApp +44 7749 776805