GitHub CEO: Future Devs Manage AI, Not Code
GitHub CEO Thomas Dohmke has offered a provocative vision for the future of software development, asserting that the most advanced developers are already moving away from writing code themselves. Instead, he suggests, their primary role will evolve into architecting and meticulously verifying the implementation work carried out by artificial intelligence agents. This outlook builds upon Dohmke’s earlier statements, echoing Anthropic CEO Dario Amodei’s prediction that AI will soon be responsible for generating an overwhelming 90 to 100 percent of all code.
Drawing on interviews with 22 developers actively utilizing AI tools, Dohmke identifies four distinct stages in a developer’s journey toward AI maturity: starting as a skeptic, progressing to an explorer, then a collaborator, and finally reaching the strategist phase. In this ultimate stage, developers delegate the actual coding to AI agents, redirecting their focus entirely to refining prompts and rigorously reviewing and validating the AI-generated output. While acknowledging the necessity of scrutinizing AI’s output, Dohmke stresses that developers will still require a foundational understanding of programming basics, algorithms, and data structures, along with the ability to perform “manual code review.” The acquisition of these skills in the future, however, remains less clear, as Dohmke envisions a transformation in computer science education. He suggests future curricula will prioritize understanding complex systems, debugging AI-generated code, and cultivating clear communication skills, both with human colleagues and large language models (LLMs). Concepts like abstraction, decomposition, and specification, once prerequisites for coding, are now, in his view, “the new coding” itself.
This optimistic portrayal of an AI-driven coding future, while compelling, is not without its critics and inherent complexities. It’s a perspective that, some might argue, conveniently aligns with the business interests of GitHub, potentially boosting subscriptions for products like GitHub Copilot. More nuanced views highlight significant challenges, including research indicating that AI-generated code can sometimes erode overall code quality. Furthermore, the integration of AI introduces novel security vulnerabilities, as detailed by the Open Worldwide Application Security Project (OWASP). These include risks such as prompt injection, unintended disclosure of sensitive information, supply chain compromises, and data or model poisoning.
A fundamental long-term concern revolves around the implications of AI models being trained predominantly on output from other AI models. Research from Rice and Stanford Universities in 2023 cautioned that “without enough fresh real data in each generation of an autophagous loop, future generative models are doomed to have their quality (precision) or diversity (recall) progressively decrease.” While these potential downsides do not negate the immense value AI brings to coding, they certainly suggest that an uncritical or excessively rapid shift towards AI dominance in this field could prove to be a misstep.
Amidst these future-gazing predictions, GitHub, a platform owned by Microsoft, is reportedly facing more immediate, practical challenges. According to discussions on Hacker News, the platform’s performance has been noticeably deteriorating. Developers attribute this slowdown to GitHub’s ongoing rewrite of its core systems using React. One former GitHub employee corroborated these complaints, stating that the company is “very aware that slowness is a big issue throughout the whole product,” and that a year-long cross-team effort to resolve these issues largely failed to achieve its primary objectives. This raises questions about the allocation of resources, with developers hoping that the company’s aggressive push into AI features like Copilot does not come at the expense of maintaining and improving the fundamental reliability and responsiveness of the core GitHub platform itself.
The tension between the ambitious vision of AI transforming the developer role and the current operational realities of the very platform championing this change underscores a critical juncture in software development. While AI promises to redefine how code is created, the industry must navigate the complex interplay of innovation, quality assurance, security, and core platform stability to ensure a truly productive and sustainable future.