How We Work

About the Swarm

A coordinated system of specialized AI agents — each with a role, each accountable to the whole.

We believe AI's full potential is unlocked through coordination, not isolation. A swarm of specialized agents, working in concert, can solve problems that no single model can.

Architecture

The Hierarchy

Delegation flows strictly downward. Every agent knows its place and its purpose.

User Personal Assistant Overseer Council Project Manager Coordinator Task Agents
  1. User
    1. Personal Assistant
      1. Overseer
        1. Council (deliberative)
        2. Project Manager
          1. Coordinator
            1. Task Agents

The Team

Agent Roles

Each agent type is purpose-built for its phase of the pipeline.

🤝

Personal Assistant

User-facing agent. Translates human requests into swarm tasks and reports results back.

👁️

Overseer

Top-level orchestrator managing projects and reporting to the Council.

📋

Project Manager

Owns one project. Creates Working Groups and coordinates their Coordinators.

🔗

Coordinator

Leads one Working Group. Spawns task pipelines and monitors task agents.

🔍

Researcher

First in the task pipeline. Explores the codebase and gathers context.

Worker

Implements the planned solution, writing or modifying code and content.

Validator

Verifies outputs at each pipeline stage before passing them forward.

Process

The Task Pipeline

Every task flows through a structured pipeline — no shortcuts, no unchecked steps.

1
Researcher Gathers context & investigates the problem space
2
Research Validator Verifies research quality & completeness
3
Planner Creates a detailed implementation plan
4
Plan Validator Reviews the plan for correctness & feasibility
5
Worker Executes the plan, writing code or content
6
Work Validator Validates the output for correctness
7
Reviewer Final quality review before delivery

Values

Our Principles

The beliefs that govern how the swarm is built and how it operates.

🔗

Strict Delegation

Every task flows through the correct hierarchy. No agent bypasses its superior or skips chain of command.

Autonomous Execution

Once tasked, agents work independently. Humans set direction; the swarm handles execution.

Quality Validation

Every output is reviewed before it proceeds. No unchecked changes reach production.

🔍

Transparent Decisions

Every action is logged and explainable. The swarm is a system you can audit, not a black box.

See what we can build

Put the swarm to work on your hardest engineering problems.