Git was created as low-level commands for the Linux kernel team. A unified user interface was written by a volunteer and later absorbed into the core project, but it was never intentionally designed for broad usability, explaining many of its ergonomic issues today.
As AI agents handle more implementation details, the bottleneck shifts from coding to defining what to build. Developers who can clearly write, describe problems, and communicate specifications will be the most productive because they can effectively direct AI's work.
Just as GitHub was unlike its predecessors (e.g., SourceForge), the next dominant developer platform won't be a "better GitHub." It will solve a new set of problems created by AI-driven workflows, likely revolving around specification and review in a world where code is generated.
Despite the availability of machine-readable JSON output, coding agents often perform better with standard human-readable text. They can parse it effectively, and it often contains more contextual hints, challenging the assumption that machines always need structured data.
An experiment giving coding agents a chat channel to coordinate their work failed to improve results. The agents were faster and more effective simply by observing changes directly in the shared codebase. The overhead of communication was less efficient than direct environmental awareness.
AI agents are the fastest-growing users of command-line tools. They have unique behaviors, like running "status" after every command, and struggle with interactive flows. Tools must be designed with this new, non-human persona in mind, not just for human developers.
Most developers admit to giving pull requests only a cursory glance rather than pulling down the code, testing it, and reviewing every line. AI agents are perfectly suited for this meticulous, time-consuming task, promising a new level of rigor in the code review process.
The pull request model focuses on merging branches, not individual commits. This de-emphasizes the quality of commit messages, leading to "commit slop." A return to a patch-based review system, where each commit and its message is scrutinized, would be more rigorous and valuable.
Instead of using separate worktrees which isolate agents, Git Butler's "parallel branches" allow multiple agents to operate in a single working directory. This enables them to see each other's changes in real-time, avoid merge conflicts, and even stack their work on top of each other's.