Software engineering is undergoing a disruptive phase of greater automation owing to the emergence of large language models (LLMs) that generate and edit code. This progress creates public excitement about AI software engineers, which promise to largely automate many core software development tasks, potentially saving tremendous costs.5 While AI-enabled code generation and code editing are now prevalent in integrated development environments (IDEs), fully automated AI software engineers are not yet widely deployed in industrial practice. What is holding back people from adopting AI software engineers? A recent blog post by the behavioral scientist and future-of-work advocate Lindsay Kohler points out that the key barrier to AI adoption is trust.3 Users are wondering if they can trust AI, and how they can demonstrate trustworthiness to stakeholders. In the domain of software engineering, the concern is thus not about the management of an organization not accepting AI software engineers, but it is about developers not trusting their new AI companions. This brings us to the question: What is the place of AI software engineers in future development workflows? If we can determine how automatically generated and manually written software can co-exist, this may give us a pathway of greater deployment of AI in software engineering! Starting from early programs of just a few lines written in high-level languages in the 1960s and 1970s, the size of programs has increased greatly to hundreds of millions of lines of code. For the past 50 years, there has been a steady interest toward programming in the large. With the increased use of AI code generation, we believe that the emphasis will be not only on programming at scale, but increasingly on programming with trust.
2026-04-01
2026-04-30