Managers can stay effective by writing small, high-impact code pieces using AI assistants, focusing on low-interrupt, high-value tasks and applying strict rules to avoid time-sensitive or low-value work.
The piece argues that a manager can still code productively by treating it as a series of tiny, high-impact contributions that fit into the gaps between meetings. The author logged 104 pull requests in six months-more than a decade of work combined-showing that disciplined, low-disruption coding is feasible even for a CTO overseeing a 50-engineer org.
Writing code as a manager often feels like a lower ROI than hiring or strategic planning, but it remains valuable for building a mental model of the codebase and making rapid, low-cost decisions. The author describes using AI assistants such as Claude Code and OpenAI Codex to answer code-base questions, generate pattern-consistent snippets, and iterate in minutes, turning what used to be weeks of work into quick, focused bursts.
To keep coding from harming the team, the author follows a tight set of rules: never touch truly time-sensitive work unless it can be finished end-to-end that day; prioritize hard-to-reach but valuable tasks like technical-debt cleanup or missing instrumentation; occasionally own a strategic project with no clear owner; and hold a higher bar for release quality, monitoring, and bug-fixing. These guardrails ensure that coding effort adds net value rather than creating hidden costs.
The upshot for technical leaders is clear: by leveraging AI tools, carving out micro-blocks of focused time, and applying disciplined selection criteria, managers can stay technically sharp, improve decision speed, and contribute code that genuinely helps the team without derailing their primary leadership responsibilities.
Part of blog:
Irrational Exuberance→Check out the full stdlib collection for more frameworks, templates, and guides to accelerate your technical leadership journey.