Kent Beck warns AI won’t replace coders, bad people skills will

Kent Beck’s blunt point is that AI is not only reshuffling coding tasks. He argues it is also changing what “good engineering work” looks like, and that exposes a weakness many teams have tolerated for years. On a recent episode of “The Pragmatic Engineer,” Beck said, “We’re kind of assholes, sometimes,” describing a pattern he says engineers often have: weaker emotional regulation, less natural empathy, and communication that can be more direct than others can handle.

Why that matters right now is simple: as AI starts writing more code, companies increasingly want engineers to review, direct, and manage AI-generated output rather than produce every line themselves. In other words, the work does not disappear, but it migrates toward coordination and accountability. That shifts the career bottleneck from “can you code” to “can you work effectively with people while steering AI-assisted production.” If you’re a product team, hiring manager, or executive, it’s a staffing and org-design issue. If you’re an engineer, it’s a survival issue.

Beck’s argument is not that engineering talent suddenly stops being valuable. Software engineers remain among the most prized talent at tech companies. The twist is that AI changes the internal labor math. When AI writes more code, it can reduce the proportion of work that is pure implementation and increase the proportion that is oversight, integration, and decision-making. Those activities are rarely solitary. They require talking to humans, aligning on tradeoffs, and making sure the right expectations are set for the right stakeholders. Beck’s framing is that engineers who lack empathy and emotional regulation will struggle more once their job becomes more people-heavy.

There is also a workflow shift that makes this kind of collaboration unavoidable. Vibecoding has become commonplace, letting seasoned coders build prototypes faster and enabling non-coders to turn ideas into pilots. That blurs the traditional boundary between engineering and product work. Instead of the clean handoff from product to engineering, AI-assisted workflows can put engineers and product teams closer together, and faster iteration pressures everyone to communicate in near-real-time.

Beck’s comments connect directly to what Business Insider reported from Anthropic’s side of the market. Anthropic’s head of growth, Amol Avasare, told Business Insider that engineers using tools like Claude Code are seeing productivity increase by two to three times. That speed bump has consequences: it puts new pressure on product managers and designers, because the engineering iteration loop tightens. When engineering can move faster, product and design have to keep up, which changes both timelines and responsibilities.

The second-order effect executives should notice is that “more speed” can produce “more interface” between functions. As AI reduces the time to draft software, organizations still need humans to define the problem, choose the direction, manage stakeholders, and decide what is acceptable risk. For smaller projects, Avasare said Anthropic is already asking engineers to act as “mini PMs.” In that model, engineers are not just responsible for the code. They also take on stakeholder coordination and cross-functional work. That hybrid “product engineer” role is basically the operational answer to Beck’s critique: technical work alone is no longer the whole job.

Beck even described the need to learn people skills as a “cosmic practical joke.” He said programmers are often told at the start that all they need is to learn everything about the computer, and they’ll be fine. Then he adds the part engineers may not expect: the human side gates your ability to affect change in the world, through communication and empathy. In practical terms, that means AI can accelerate the “making” portion of software, but your career and your team outcomes still depend on the “making sense together” portion. That includes regulating conflict, understanding what others need, and translating between technical execution and human expectations.

So the strategic stake for executives and investors is not whether AI will write code. It will. The stake is whether your teams are built to absorb that shift. If engineers are increasingly asked to review AI output, direct how it should behave, and coordinate across functions, then your org design and talent development strategy need to treat people skills as part of engineering excellence, not as a soft add-on. In the AI boom, the most vulnerable roles are not the least technical. They are the least adaptable at the human layer, where trust, clarity, and emotional regulation decide what gets shipped and what gets stuck.