A developer just published a short essay with a observation that anyone who's spent time pair-programming with an AI will recognize immediately: working with AI feels more like leadership than coding.
The post, which appeared on Hacker News, articulates something that's been an unspoken undercurrent in the AI coding tools conversation. When you work with Claude, Copilot, or ChatGPT on a coding task, the skills you're using aren't the skills you learned in computer science class. They're the skills of management.
The Leadership Parallel
Think about what a good engineering manager does:
- Delegates tasks with clear specifications
- Reviews work for correctness and quality
- Provides context about the broader goals and constraints
- Gives feedback when the work isn't quite right
- Breaks down complex initiatives into smaller, well-defined tasks
- Knows when to step in and when to let the team figure it out
Now think about what you do when working with an AI coding assistant:
- Write prompts that specify what you want built
- Review generated code for bugs, security issues, and style
- Provide context about the project, conventions, and requirements
- Iterate with feedback when the output isn't right
- Break complex features into smaller, AI-manageable chunks
- Know when to write it yourself vs. when to let the AI try
The overlap is uncanny. And it's not coincidental.
Why This Shift Matters
This reframing has practical implications for how we think about developer skills:
Code review becomes the core skill. When the AI can write the code, your value shifts to evaluating it. This means deep understanding of correctness, security, performance, and maintainability — the same skills a senior engineer uses in code review, not the skills they use in writing code from scratch.
Specification becomes a superpower. The developers who get the most out of AI tools aren't the ones who write the best code — they're the ones who write the clearest specs. This is literally what product managers do. The skill of "describe what you want precisely" is undervalued in traditional software engineering but becomes critical when your "team" is an AI.
Context management is the new architecture. A manager doesn't tell a senior engineer how to implement a function. They provide context about the business goals, constraints, and priorities. Similarly, the most effective AI-assisted development involves providing rich context — project conventions, existing patterns, edge cases to consider — rather than line-by-line instructions.
The Skills Gap
Here's the uncomfortable truth: most developer education and career progression focuses on implementation skills. We learn algorithms, data structures, framework APIs, and design patterns. We measure seniority by the complexity of systems we can build.
But if working with AI is more like leadership, then the skills that matter most are:
- Clear written communication — can you describe what you want unambiguously?
- Systems thinking — can you see how individual pieces fit into the whole?
- Critical evaluation — can you quickly assess whether code is correct, secure, and maintainable?
- Strategic decomposition — can you break a large goal into well-defined, independently verifiable tasks?
- Feedback calibration — can you give precise, actionable feedback that improves the next iteration?
These are leadership skills. And they're not typically taught in CS programs or bootcamps.
What This Means for Teams
For engineering organizations, this shift has implications:
Pair programming with AI changes team dynamics. When every developer has an AI assistant, the gap between junior and senior developers might narrow on implementation but widen on specification and review. Junior developers who can write good specs and evaluate output will accelerate faster than those who rely on the AI to do their thinking.
Code review becomes more important, not less. If AI generates more of the code, human review is the last line of defense. Investing in review skills — security, performance, correctness — becomes critical.
The IC track might start looking like the management track. If the day-to-day work of an AI-assisted developer resembles management more than implementation, the skills and career paths may converge. This could be disorienting for developers who chose IC specifically to avoid management work.
Conclusion
The insight that working with AI feels like leadership isn't just a clever observation — it's a signal of where the profession is heading. The developers who thrive in the AI era won't necessarily be the ones who write the best code. They'll be the ones who can direct, review, and coordinate AI outputs the way a good leader directs a team. That's a different skill set, and it's one we should be deliberately cultivating.
Top comments (0)