← Series One

Old Ideas, New Consequences

What the thinking we already have can teach us about the technology we're building.

A new tool can make an old problem harder to recognize. Goodhart’s law helps explain why a system can meet a target while defeating its purpose. Tacit knowledge helps explain why a workflow documented on paper can fail once you remove the people who know when to bend its rules. This strand takes ideas from economics, philosophy and systems thinking and asks where they still hold when machines can reason and act at scale.

Why it matters

Recognizing the old mechanism changes what you look for before a rollout. A slow handoff might catch exceptions the dashboard never sees. A cheaper way to produce work might leave the people who review it with a larger queue. Each analogy has limits. The test is whether it helps predict who gains, what breaks, and which safeguards are worth keeping.

Published essays

  • Defending Friction

    Before eliminating enterprise friction, understand what it protects. Apparently inefficient processes may contain safeguards nobody knew existed. This is the completed opening essay.

Planned stories

  • When the Metric Eats the Mission

    AI can optimize performance indicators relentlessly while undermining the underlying purpose those indicators were intended to represent.

  • The Knowledge We Didn't Know We Had

    Organizations depend on undocumented judgments and habits. Automating formal workflows may destroy the unwritten expertise that made them work.

  • The Machine That Gave Us More Work

    Cheaper production can increase demand. AI may multiply the code, documents and proposals that humans must evaluate rather than straightforwardly shrinking workloads. This is a possibility, not an inevitability.

  • The Map That Ate Reality

    Models simplify reality. When AI-generated abstractions dominate organizational decisions, what information becomes invisible or actively ignored?

  • When Everything Looks Like an Agent

    A powerful technology can turn into a universal hammer. Some problems need deterministic software, better interfaces or process change instead of autonomous AI.

  • The Catastrophe Nobody Caused

    In complex tightly coupled systems, failures may emerge without a single broken component. Connected AI agents intensify the importance of dependency analysis and containment.

  • The Information Nobody Can Centralize

    Relevant information is distributed and context-dependent. Centralized intelligence may miss what local participants know and how conditions are changing.

Other perspectives