Essays

Red Teaming AI Systems Is Not Optional

Only 4 of 23 AI systems evaluated had undergone adversarial testing. Untested systems averaged 3.7 exploitable vulnerabilities including prompt injection and data extraction paths.

The latent skills hypothesis: What we lose when we stop doing things the hard way

The Data Engineer’s Guide to Cost-Aware Architecture

Cost-aware architecture patterns reduced monthly cloud data spend from $14,200 to $6,800 without degrading query performance. Five techniques every data engineer should apply.

Cognitive load theory applied to AI interface design

Configuration as a First-Class Architectural Concern

Configuration-related failures caused 43% of production outages across 19 systems, exceeding code bugs at 31%. Configuration deserves the same design rigor as source code.

Secular Transcendence in Technical Work

Technical work can provide genuine transcendence: flow states, elegant solutions, and architectural beauty. These experiences share the structure of religious experience without requiring its metaphysics.

The Hidden Cost of Convenience Architecture

Convenience architecture patterns (magic defaults, implicit behavior) caused 41% of critical incidents across 23 systems, with diagnosis taking 3.7 times longer than explicit code failures.

What Aristotle Would Say About Algorithmic Virtue

Aristotle would not ask what the algorithm should do. He would ask what kind of engineers we are becoming by building it. The gap between compliance and character is where algorithmic harm lives.

The attention ecology: Understanding focus as a limited resource in a networked environment

The Ethics of AI-Generated Content at Scale

AI systems generate an estimated 15% of all web content daily. When content production outpaces human evaluation, the epistemic environment degrades for everyone.