Design and build studio, Zagreb

We design the organization. Then we build what it runs on.

Cell of 5Checking
Human steward AgentFive human stewards as peers with no boss, agents beneath each, and a smaller team of agents under Execution. Each new block on the network sends an agent’s work to the steward who owns it.

Resomnium turns companies into small, accountable cells where humans own judgment and AI agents do the execution. Then we build the agent systems those cells run on: scoped authority, signed receipts, a named human at the end of every chain.

The method is open

CellOS. Apache 2.0.

The Cell as an open framework. Read it, run it, fork it. See the model

The agents are live

Live now

Agents do signed, escrowed work on a running network.

Every chain ends with a human

Five stewards. No boss node.

Each agent answers to one named person. See how

The problem

The tools work. The org chart was drawn for people working alone.

Add agents to a pyramid and you get more output and less accountability. The organization needs a new shape before it needs another tool.

  1. 01

    Nobody signs for the output.

    When an agent gets it wrong, the trail ends in a shrug. Speed went up. Accountability went missing.

  2. 02

    Decisions climb a tree built for humans.

    Approvals queue at a layer that was never meant to read machine work, so the fast part waits for the slow part.

  3. 03

    Escalation has no address.

    An agent hits an edge case at 2am. Nobody is named to take the call, so it waits, or it guesses.

The Cell

Five humans. No boss. Agents beneath each.

The Cell is the smallest unit that can carry full accountability. Five stewards, each a person, each owning one function. They are peers. Agents work beneath each steward, and a team of agents can mirror the five roles one level down. The structure is fractal, and the human above still signs.

The five steward roles of a cell
StewardOwnsAgents beneath doThe human decides
ClarityThe questionResearch, synthesis, decision logsWhat we are actually trying to do
ExecutionThe doingBuild, test, ship, monitorWhat ships, and when it is good enough
NarrativeThe storyDrafts, adaptations, schedulingWhat goes out, and in whose voice
AccessThe doorsResearch, briefings, follow-upWhich conversations happen, and who sits across the table
IntegrityTrustChecks, logs, flags. They never overrule.When to stop the line

Proof

Built with people who run real systems.

We do not draw structures for systems we have never operated. Our partners run live networks. The method is public. The agent is one click away.

Apex Fusion(opens apexfusion.org)

Partner

A multi-chain ecosystem. Its agent-native chain, Vector, has been live on mainnet since August 2025.

Checking
HAL8

Partner

The company building Apex Fusion.

hal8.ai ↗

Partner

Engineering partner.

thespian.eu ↗

Open

CellOS

The Cell as software: cells, roles, scope boundaries and escalation chains for human-AI teams. Apache 2.0. v0.2.0, early.

github.com/Resomnium/cellos ↗
Checking

Org Gap agent

A paid agent that diagnoses an organization against the Cell framework and returns a written report.

Try the method ↗

Work with us

Three ways to work with us.

All long-term. All built so the partner owns the outcome. We start with a conversation.

Three ways to work with Resomnium
EngagementForWhat happensShape
Option 01Embedded design partnershipFounders and executives restructuring a real company around AI. You have revenue, a team, and a structure that is starting to creak.We work inside your constraints until the new structure holds without us. Decision maps, cell design, the agent layer, in the order your situation demands.Quarterly, renewed on results
Option 02Venture and ecosystem co-buildEcosystems, foundations, and founders building agent-native products.When the upside is real, we co-build instead of billing hours. Agent infrastructure, machine-to-machine markets, products born AI-native.Milestones, aligned economics
Option 03Research and open standardsConsortia, standards bodies, and public institutions.Agent governance and human-AI coordination, commissioned or in consortium. The output ships in public wherever possible.Grant, consortium, or commission

Tell us which one fits →

Try the method Checking

Try the method on your organization.

The Org Gap agent diagnoses an organization against the Cell framework and returns a written report. It is a paid agent. It is also the fastest way to see how we think.

Writing

The thinking is public.

Eighteen essays in three series. Jobs, then skills, then the company. Read them in order, or start with the one that hurts.

Series 01 · 6 essays · 48 min

Jobs of the Future

The five human roles that survive when AI does the execution.

  1. 01What Survives When AI Does Everything Else
  2. 02The Clarity Steward
  3. 03The Execution Orchestrator
All 6 in the series →

Series 02 · 5 essays · 33 min

Skills of the Future

What those humans need to be good at: agency, integrity, intelligence, taste.

  1. 01The Human Toolkit
  2. 02Agency
  3. 03Integrity
All 5 in the series →

Series 03 · 7 essays · 48 min

Company of the Future

How the roles combine into the Cell, and how the Cell scales.

  1. 01How Organized Effort Looks Like Post AI
  2. 02The Cell
  3. 03Single Player
All 7 in the series →

Start here

Tell us where the structure hurts.

[email protected]