The post says coding with AI feels more like leadership because you spend less time typing code and more time framing work, steering execution, and evaluating results. Most people bought the underlying observation but rejected the branding. The stronger reading was that agentic coding feels like management, tech lead work, product specification, or orchestration with the human side stripped out. You still need vision, decomposition, and communication. You do not need trust building, motivation, morale management, or any of the social friction that makes leading people hard.
That distinction mattered because a lot of the conversation landed on what skills actually transfer. People with management experience said they are comfortable delegating imperfectly, waiting to see what comes back, and iterating on instructions instead of trying to preprogram every step. Others said that is overstated. Real leadership involves autonomy, disagreement, incentives, careers, and consequences. An
LLM has none of that. It is closer to an infinitely available junior contractor, outsourcing shop, or code generation machine that can move fast, miss subtleties, and confidently insist it is right.
The practical consensus was sharper than the semantic fight. Good results come from bounded tasks, explicit requirements, acceptance criteria, and strong review. Several experienced users said they still feel like engineers, not managers, because the core work remains architecture, separation of concerns, debugging, and code review at higher volume. The people generating the most slop are the ones treating models as autonomous builders or as authoritative experts. That failure mode looks especially bad when nontechnical leaders use
Claude as a substitute for engineering judgment and then dump the cleanup on actual developers.
A second theme was labor structure. Some people are already freezing hiring because AI lets experienced operators absorb more work. Others warned this creates a pipeline problem for junior developers and amplifies weak technical leadership by letting people produce large volumes of plausible but brittle code. The mood was not anti-AI so much as anti-anthropomorphism and anti-hype. The useful frame was not "AI teaches leadership." It was that AI turns more software work into delegation, specification, and verification, while making taste and technical oversight more valuable, not less.
allen.bargi.org