At the Thoughtworks Future of Software Development Retreat, industry leaders concluded that AI-assisted development is accelerating existing problems rather than solving them. The consensus was clear: without established software delivery practices, AI's velocity multiplier becomes a debt accelerator. Rachel Laycock emphasized that AI amplifies whatever infrastructure exists, making poor practices worse while potentially improving good ones.

Key themes emerged around the need for new engineering disciplines. The retreat identified risk tiering as the new core engineering discipline and TDD as the strongest form of prompt engineering. Participants discussed how LLMs are eating specialty skills, potentially reducing the need for specialist front-end and back-end developers as LLM-driving skills become more valuable than platform-specific knowledge. The human side of development is shifting rapidly.

Security concerns dominated discussions, with one enterprise noting they deliberately stay a quarter behind the leading edge to manage risks. Platform thinking emerged as essential, with teams needing to create "bullet trains" for safe AI adoption. As Adam Tornhill's research showed, AI performs 30% better in healthy codebases, highlighting that technical debt directly impacts AI effectiveness. The industry faces a critical choice: either establish robust practices now or risk amplifying existing problems at unprecedented scale.