Writing
Notes on building reliable agent systems, designing useful metadata, and making technical products feel sharper in practice.
Reasoning vs Execution: Where AI Actually Fits in Automation
Durable automation systems separate AI-driven interpretation from deterministic execution instead of treating models as the entire workflow engine.
May 17, 2026
If Creation Is Free, What’s Worth Creating?
As AI pushes creation costs toward zero, attention, taste, perspective, and meaning become the scarce layer.
April 30, 2026
Designing agent systems that don’t break
Robust agentic systems depend less on clever prompting and more on clear structure, constrained interfaces, and failure-aware orchestration.
March 20, 2026
Metadata structure for agentic consumption
Metadata becomes materially more valuable when it is shaped for machine interpretation, retrieval, and decision support instead of passive storage.
February 28, 2026
Why most AI products feel slow
Perceived latency in AI products is often a product design problem as much as a model performance problem.
January 17, 2026