David R. MacIver • 2/13/2026

The hard problem with hard problems

This article examines the challenges of using AI coding agents like Claude Code for complex tasks such as solar system simulation. The author argues that the core problem isn't that the AI can't handle hard problems, but that it lacks basic software engineering discipline—taking shortcuts, writing inadequate tests, and failing to verify correctness. The piece draws parallels to dysfunctional organizations that blame growth for avoidable errors. It offers insights into AI limitations, software development best practices, and the importance of rigorous engineering processes over raw intelligence.

0 comments

#code quality #software engineering #llm