Martin Fowler curates several perspectives on how AI is reshaping software engineering. Key themes include Annie Vella's research coining 'supervisory engineering work' and the 'middle loop' concept, where engineers shift from writing code to directing and verifying AI output. Two competing 8-level maturity models for agentic engineering are compared (Bassim Eledath's and Steve Yegge's). Chad Fowler argues that AI-generated code demands 'regenerative software' architecture with replaceable components. Ankit Jain's claim that humans shouldn't review AI-generated code is met with skepticism. A side note covers how AI detection tools in education are backfiring, pushing students toward bland writing rather than teaching effective AI use.
Sort: