The Journal
Essays
Long-form writing on AI, philosophy, psychology, and systems thinking.
-
Bias Detection Tools Are Only as Good as Your Data Model
Bias detection tools identified only 34% of bias issues I encountered in production. The other 66% originated in data modeling decisions upstream of any tool's reach.
-
AI Ethics Certification Is Credential Theater Without Engineering Practice
AI ethics certifications test knowledge of principles but ethical AI requires engineering discipline. I found zero correlation between certification and practice quality across 9 organizations.
-
Event Sourcing as Organizational Memory
An event-sourced system maintained 2.8 million state changes over 3 years, enabling state reconstruction at any point in time and reducing audit preparation by 83%.
-
Ethics of AI in Hiring: Algorithms That Gate Opportunity
Four AI hiring systems embedded discriminatory patterns through architectural choices affecting 2.3 million applicants annually. The discrimination was not intentional. It was architectural.
-
Emotional AI and the Boundary of Machine Perception
Emotion detection accuracy ranged from 62-78% for basic emotions and 31-45% for complex states across 3 systems. Classifying surface patterns is not perceiving emotions.
-
On Building Things That Last in a Culture of Disposability
We live in a culture of disposability. The average system is deprecated within 7 years. Wabi-sabi and Stoic durability offer a counter-philosophy: build for endurance, not novelty.
-
Python’s Gravity Well: Language Choice Shapes Architecture
Python is present in 92% of data pipeline codebases, creating path dependencies that constrain infrastructure for years. Its gravity well requires strategic, not revolutionary, escape.
-
Data Privacy Engineering Is a Data Engineering Discipline
Implementing tokenization and differential privacy at the pipeline level reduced PII exposure incidents by 89% while adding less than 3% to processing time.
-
The OKR Trap: Why OKRs Break at Scale
OKRs created 312 person-hours of overhead per quarter at a 120-person org. The framework fails when used to create alignment rather than represent it.
-
The architecture of a second brain that actually works