High-throughput Atlassian engineers, identified through PR throughput data and peer nomination, work on brownfield codebases and find that AI accelerates their long-standing fundamentals rather than making them obsolete. Key practices include small, single-responsibility PRs, which are easier for humans and agents to reason about and less prone to merge conflicts, with faster builds lowering the cost of managing many small changes. Spec-driven development, an old discipline, now serves as the context agents execute against, with AI-native frameworks like BMAD repackaging it for agents. Engineers break projects into tightly scoped Jira work items, assign them to coding agents running in parallel, and review draft PRs, using the work item to coordinate atomic units of work and minimize overlap. Richer specs with repo context, code paths, acceptance criteria, test levels, and gating produce more reliable agent output, so engineers with the "write the spec first" habit benefit most from AI. Structured intent outperforms ad-hoc prompting, and the gap between engineers who apply these fundamentals and those who don't has become more noticeable in an AI-accelerated world.