Writing
-
Engineering Natural Language Interfaces: Compiling Intent
A practitioner survey of approaches for compiling natural language into validated commitments — classical pipelines, constrained decoding, structured outputs, logprob overlays, and judge-based confidence mechanisms, with a harness discipline that makes technique comparisons engineering instead of preference.
-
The Future of Software May Be Conversational Rather Than Autonomous
The most durable role for LLMs may not be replacing deterministic systems — it may be reducing the friction between people and those systems.
-
Conversational Software Engineering: Compiling Intent
A concrete approach to LLM-assisted engineering: conversation as the control surface, specifications compiled into a validated artifact, and the compiled spec as the center of gravity everything else derives from.
-
AI Didn't Simplify Software Engineering: It Just Made Bad Engineering Easier
AI tools lower the barrier to producing code, not to building reliable systems. The hard part — maintaining alignment between specifications, tests, and implementation — has not gone away. Expertise still matters.
-
Engineering Alignment in Probabilistic Generation
A structural model for achieving and maintaining correctness in LLM-assisted systems, grounded in the construction of a production financial planning system spanning ten modules and nearly 150 specifications.
-
Governing Correctness in LLM-Assisted Development
LLMs don't break software correctness — they relocate where correctness must be governed. A practitioner account of boundary drift, interpretive boundaries, and the discipline required to prevent silent divergence.