While agentic coding offers undeniable utility, it is also introducing significant challenges to the craft of software engineering. Alex Martsinovich identifies four primary concerns, which he terms the "four horsemen of agentic coding."
The Four Horsemen of Agentic Coding: Four Negative Impacts of AI Agents in Software Engineering
First, LLM-generated code is increasingly characterized by "slop"—a term for code that, while functional, lacks human-like quality and can become incomprehensible or "unemployable" in its style. This can transform shared codebases into "AI wastelands" that engineers no longer feel comfortable maintaining.
Second, agentic coding increases the distance between engineers and their work. As developers shift from direct production to overseeing agents via vague instructions, they may experience a sense of detachment and a diminished sense of ownership over the final product.
Third, the use of AI is contributing to "deskilling." The ease of using AI agents reduces the incentive to undergo the traditional, difficult learning curves required to master software engineering, potentially disrupting the long-term supply of skilled professionals.
Finally, the rise of autonomous agents is weakening the social fabric of engineering teams. As developers spend more time interacting with AI agents rather than collaborating with human colleagues, interpersonal communication and the ability to express individual expertise through code are being diminished.
Sources
- The Four Horsemen of Agentic Coding （Hacker News Frontpage, 2026-10-02）
- Alex Martsinovich