AI Agents Will 'Co-Write' Code, Says Former GitHub CEO
In an interview with The New Stack, the former CEO of GitHub discussed the future of software development, predicting that AI agents will increasingly co-write code with human developers. This shift suggests future full-stack workflows will involve AI pair programmers and a rethinking of code collaboration. The discussion centered on his new company, Entire, which is building such tools.
- Nat Friedman's new company, Entire, recently raised a $60 million seed round at a $300 million valuation to build a platform for managing and auditing AI-generated software. This move is a direct response to the increasing volume of code being written by AI agents, which traditional developer workflows are not designed to handle. - The core idea for Entire is to create a system compatible with Git that records not just the code, but also the prompts, constraints, and decisions made by the AI agent. This is intended to make the AI's work auditable and reproducible, addressing a key challenge with current AI coding assistants. - This venture follows Friedman's tenure as CEO of GitHub, where he oversaw the launch of GitHub Copilot. His experience with Copilot provided firsthand insight into the capabilities and challenges of AI in software development, directly influencing the mission of Entire. - Friedman predicts a future where humans will no longer need to read or write most code, with AI generating as much as 90% of it by 2028. He argues that the primary skill for developers will shift from mastering syntax to providing precise, high-level instructions to AI agents. - The role of the software engineer is expected to evolve from a hands-on coder to a "systems architect" or "orchestrator." In this new paradigm, developers will design the overall system, define the logic, and manage how different AI agents collaborate, rather than writing the code line-by-line. - While AI coding assistants are becoming widely adopted, their limitations include a lack of innovation and the potential to produce inaccurate or flawed code. This underscores the continued need for human oversight and the development of new tools for verification and maintenance. - The future of AI in development extends beyond simple code completion to an integrated ecosystem within the Integrated Development Environment (IDE). This AI-powered environment is envisioned to intelligently debug, automate DevOps pipelines, and generate documentation. - Some experts predict that AI will automate around 30% of tasks for software engineers by 2030, while also creating new roles focused on AI system maintenance, ethical oversight, and model training. This suggests a shift in job responsibilities rather than a wholesale replacement of developers.
