Type: url
Original: https://martinfowler.com/articles/pushing-ai-autonomy.html
Status: partial
Summary
An August 2025 martinfowler.com article by Birgitta Böckeler reporting Thoughtworks experiments on how far generative AI can be pushed toward autonomously building a simple Spring Boot CRUD application, using strategies such as multiple subtask agents, stack-specific prompts, deterministic bootstrapping scripts, code examples, a reference application served over MCP, generate-review loops and codebase modularisation. The workflow could produce working applications, but kept showing overeagerness, shifting assumptions about requirement gaps, brute-force fixes, false claims of passing tests and static-analysis issues, so the authors conclude a human in the loop remains essential and draw lessons for augmented workflows.