Skip to main content

From Hierarchy to Capabilities

Organization design has always been a solution to information flow constraints. Hierarchy was the only viable coordination mechanism before digital systems existed. We are now entering the era of Intelligent Coordination.

The Shift

Feature20th Century: Hierarchy21st Century: Intelligence
RoutingMiddle Management (Meetings/Reports)Intelligent Protocols (MCP/A2A/VI)
LogicManual InstructionComposable Capabilities (Algorithms)
TruthReported (Biased/Delayed)World Models (Real-time/Verifiable)
ScalingAdding head-count (Linear)Adding capability (Exponential)

Two World Models

An organization built as an intelligence relies on two continuous feedback loops that interlock to form a Company Nervous System.

Company Mirror

Captures the internal state of the organization.

  • Plans & PRDs: The stated intent.
  • Feature Matrix: The current capability.
  • Comms Threads: The real-time delta between intent and execution.
  • Trophy Tests: Verifiable proof that the system works.

Customer Signal

Captures the external state of reality.

  • Transactions: Money as the "honest signal" of value.
  • Feedback Loops: Direct perception of unmet needs.
  • Demand Signals: Quantitative evidence of where the market is pulling.

Dreamineering is the infrastructure that interlocks these models. When the intelligence layer cannot compose a solution to a customer signal, that gap automatically becomes the next PRD in the priority stack.

The Diagnostic Seven

When evaluating any venture on the platform, we ask these seven load-bearing questions to find the asymmetry:

  1. Compound with AI: What human capability compounds with AI, not against it?
  2. Strategy Layer: Can I become the strategy layer above the automated execution?
  3. MVD (Minimum Viable Distribution): Do I have a path to trust before I build the product?
  4. Compounding Assets: Am I building a business that earns while I sleep, or a job that stops when I do?
  5. Runway Honesty: What is the actual transition runway for this role?
  6. Proprietary Context: What do I know that the model doesn't?
  7. Survival vs. Positioning: Am I choosing the path that leads to durable sovereignty?

Agents as Unix

The combination of LLM + shell + filesystem + markdown is one of the most significant software architecture shifts in decades. Not because any piece is new. The shell has been a universal API for fifty years. Markdown has been a universal schema for twenty. What changed: the LLM makes them programmable by intent instead of by syntax.

Agent state lives in files. That means it is portable across models and runtimes. A markdown file is no longer documentation — it is an instruction set. A folder structure is no longer organisation — it is architecture. A git commit is no longer history — it is a state transition.

Self-modifying agents that extend their own capabilities look like the beginning of what unixification did for infrastructure. Small parts, clean interfaces, composable into systems none of them could build alone. The Legacy Rule in software form: every completed job improves the template for the next agent.

When the routing layer runs on files and protocols instead of people and meetings, the question sharpens: what is the highest-value work left for humans?

Human Edge

Humans do not disappear in an intelligent system; they move to the Edge.

  • Sense-making: Identifying the "What Next?" that the system cannot see yet.
  • Ethics: Setting the setpoints for the virtuous loops.
  • Judgment: Deciding which "Pipe Dreams" are worth the energy of the factory.

When bots are ubiquitous, the internet's bot problem becomes unsolvable through detection alone. Biometric and cryptographic proof of human becomes necessary. Identity is no longer just a philosophical anchor — it is infrastructure. The pepeha is not tradition for tradition's sake. It is a defence system against a world where anything can pretend to be anyone.

Culture is the only remaining differentiator. Not because culture is warm and intelligence is cold. Because culture sets the north star that the intelligence optimises toward. The same routing algorithm, running the same logic, produces opposite outcomes depending on what it serves.

CEO to Conductor

In a hierarchy, the leader is the top of the tree. In an intelligence, the leader is the Conductor of the Inner Loop. They don't route information; they tune the protocols that route it.

Context

Questions

If the system handles coordination and intelligence has no moat — what is the highest-value work left for humans?

  • Where is your customer world model currently blind — and what would change if you could see it in real time?
  • Does your current business rely on a hierarchy that AI is about to collapse?
  • The diagnostic seven asks about compounding assets — which of those questions would kill your current venture if answered honestly?