The Journal

Essays

Long-form writing on AI, philosophy, psychology, and systems thinking.

  • · 4 min read ·Process

    Why Automation Amplifies Process Failures

    Automation does not fix broken processes; it accelerates them. After mapping a 9,000-step workflow, 2,100 steps existed solely to correct errors from earlier steps. Automating would encode those errors at machine speed.

  • · 4 min read ·Data

    Signal Extraction in an Age of Information Obesity

    The abundance of available information has not produced better decisions. Organizations drowning in data are not those with too little information but those without a framework for deciding what information matters.

  • · 4 min read ·Architecture

    Fault Tolerance as an Organizational Principle

    Fault tolerance is an organizational principle: the capacity to continue functioning when individual components fail. Organizations most vulnerable to disruption concentrate critical knowledge in single individuals rather than distributing it.

  • · 5 min read ·Process

    The Invisible Tax of Context Switching

    Context switching costs 23 minutes of recovery per interruption, yet organizations design workflows requiring 40+ daily transitions. The invisible tax is not the switch itself but the cumulative degradation of judgment across every transition.

  • · 4 min read ·Data

    Goodhart’s Law and the Weaponization of KPIs

    Goodhart's Law states that when a measure becomes a target, it ceases to be a good measure. After managing 30+ metrics across 1,000+ programs, the metrics with highest organizational visibility were consistently least representative of actual health.

  • · 4 min read ·Architecture

    Conway’s Law as Applied Psychology

    Conway's Law states that systems mirror organizational communication structures. In practice, this mirroring operates as applied psychology: technical architecture reveals trust patterns, power hierarchies, and unresolved conflicts of the building team.

  • · 4 min read ·AI Systems

    Hallucination Is Not a Bug

    Language model hallucination is treated as a defect, but it is a fundamental property of probabilistic text generation. Models will always produce confident, occasionally fictional outputs because the mechanism enabling utility is the same one producing errors.

  • · 4 min read ·Philosophy

    Constraint as a Teacher in Software Design

    The most reliable architectures emerge not from unlimited choice but from deliberate limitation. Constraints function as teachers by forcing decisions that unconstrained environments allow engineers to indefinitely defer.

  • · 4 min read ·Architecture

    Why Every Database Migration Is a Rewrite

    Database migrations are marketed as reversible schema changes, but every migration of sufficient complexity becomes a rewrite. After migrating 15,000 records across 47 tables, 34% of the effort was reconciling business logic encoded in the schema itself.

  • · 5 min read ·Data

    The Myth of the Clean Dataset

    After processing 36,791 SEC filings, 23% contained structural anomalies that would have corrupted any downstream analysis. The myth of the clean dataset persists because most practitioners encounter data only after someone else has already cleaned it.