AI demands more engineering discipline. Not less

11 hours ago

AI engineering requires increased discipline, not less, especially with nondeterministic systems.
The shift from handcrafted server pets to immutable infrastructure parallels current AI-driven changes in software development.
In 2025, AI-generated code became as good as median human code, turning code from a precious asset into a disposable, regenerable resource.
The true product of software teams is shared understanding, but production is the ultimate reality, necessitating tests and validation in production.
AI tools can enhance engineering discipline through fast feedback loops and better artifacts for review, like architecture diagrams over code.
Phoenix Architectures advocate replacing rather than fixing code, aligning with AI's ability to regenerate code cheaply.
Human strengths like creativity and inspiration remain valuable, while validation tasks are better suited to machines.
Discipline in engineering will drive value, as determinism and user experience remain critical despite AI advancements.
The future involves using AI to handle validation and repetitive tasks, allowing humans to focus on higher-level design and problem-solving.
Ethical considerations around AI's impact on jobs and creativity are acknowledged and will be addressed separately.

Hasty Briefsbeta