Blog

Notes on engineering intelligence, metrics, and the AI-native era.

Engineering intelligence

What Karpathy's LLM Wiki Means for Engineering Leaders

Andrej Karpathy's viral LLM wiki post is being read as an argument about retrieval architecture, but the new thing in it is the role he gave the LLM. He stopped using it as a question answering machine and started using it as a long running curator of a knowledge stream. That role applies far more powerfully to engineering management than to personal research, which is the part nobody is talking about yet.

Engineering metrics

What 2026 Research Says About Engineering Productivity

94 percent of business leaders consider AI critical to their five year strategy, but 60 percent of engineering leaders say they lack the metrics to manage it. The 2025 and 2026 research converges on a few findings: gains are real but uneven, rework cost is significant, the economic unit has shifted from seats to tokens, and the strongest predictor of realized value is reinvestment of gains rather than adoption rate.