Thoughtworks Technology PodcastThoughtworksWhat does the future of software engineering look like?
In short
This podcast episode discusses the future of software engineering, drawing insights from a recent ThoughtWorks event. Guests Andrew Harmel-Law and Keith Morris highlight the shift from exploring 'shiny new tech' to practical application of AI, emphasizing the need to redefine human roles, manage trust in AI-generated code, and address the challenges AI poses to existing development practices and organizational structures. The conversation delves into the impact of AI on infrastructure as code, the concept of agent 'personalities' versus defined roles, and the enduring debate around control and quality in an AI-amplified development environment.
Key takeaways
- AI's utility in software engineering is becoming undeniable, shifting focus from novelty to practical application and addressing new challenges.
- The role of humans in software development is recalibrating, with increased emphasis on defining tasks, providing guidance to agents, and understanding where to intervene.
- AI amplifies existing development habits, exposing weaknesses in current practices and necessitating a re-evaluation of quality, governance, and control mechanisms.
- Conscious thinking about system architecture, business problems, and non-functional requirements remains crucial, even as AI handles more implementation details.
- The 'theater of control' is a recurring theme, questioning how much control developers truly have and how to manage trust in AI-generated code and decisions.
Chapters
Introduction to the Podcast and Guests
Ken introduces the ThoughtWorks Technology Podcast and welcomes Andrew Harmel-Law and Keith Morris from the UK office.
Insights from the Future of Software Engineering Event
Discussion about a recent ThoughtWorks event in Europe, held under Chatham House Rule, focusing on the future of software engineering and comparing it to a similar US event.
Shifting Focus: From Shiny Tech to Practical AI
Guests note a shift from exploring novel AI concepts to practical application, with the European event focusing on known capabilities and addressing implementation challenges.
AI's Impact on Infrastructure as Code
Keith discusses the challenges of AI generating infrastructure code, highlighting its tendency for 'silly mistakes' and the need for more directive guidance compared to application code.
Agents, Team Structures, and Conway's Law
Andrew explores how AI agents might influence team communication structures (Conway's Law), potentially blurring boundaries and raising questions about managing agent access and code base integrity.
Defining Agent Roles vs. Personalities
The conversation delves into whether agents should be given 'personalities' or clearly defined roles and expertise, with a preference for task-oriented agents and structured guidance over anthropomorphizing.
The Illusion of Control in AI-Driven Development
Discussion on the 'theater of control,' questioning the extent of human control in an AI-amplified environment and whether traditional review processes like pull requests are truly effective.
AI Amplifying Development Habits and Weaknesses
Guests explore how AI amplifies existing development practices, exposing areas where teams might have 'gotten away with' less rigorous habits, now necessitating higher standards for quality and governance.
Enduring Principles: What Survives and What Changes
Andrew and Keith reflect on what aspects of software engineering will remain crucial (e.g., conscious architectural thinking, non-functional requirements) and what practices might evolve or be re-evaluated in an AI-driven future.
Concluding Thoughts and Importance of Depth
Ken concludes the podcast, thanking the guests and emphasizing the value of in-depth learning and foundational knowledge in the rapidly evolving tech landscape.
Get the next Thoughtworks Technology Podcast recap
A short recap of every new episode, by email. Free for up to 5 shows.
Summary by InboxHiive. Not affiliated with Thoughtworks Technology Podcast. Written with AI from the episode audio; check the episode for exact quotes.