Lars Faye's article "Agentic Coding Is a Trap" has ignited discussion in the developer community about the hidden costs of over-reliance on AI coding agents. Published on his blog and gaining 73 points with 47 comments on Hacker News on May 3, 2026, the piece presents a systematic critique of how agentic coding tools may be creating long-term problems while solving short-term productivity challenges.
The Supervision Paradox Creates a Fundamental Contradiction
Faye identifies what he calls the "core paradox of supervision": developers need strong coding skills to effectively oversee AI agents, yet using these agents causes those very skills to atrophy. He cites Anthropic's own researchers who noted that "effectively using Claude requires supervision, and supervising Claude requires the very coding skills that may atrophy from AI overuse." This creates a dangerous feedback loop where the tools designed to enhance productivity may ultimately undermine the expertise needed to use them effectively.
Skill Degradation and Disconnection From Code
The article presents evidence of significant cognitive decline among developers heavily dependent on AI tools. Even experienced programmers report losing mental models of their own applications, making it harder to reason about additional features. Faye cites a 47% decline in debugging capabilities as a documented example. Traditional development prioritizes understanding, standards alignment, and code quality, but agentic coding inverts these priorities—emphasizing speed and volume over comprehension. As Faye writes, "What you say is often not what you mean, and LLMs fill in ambiguity with assumptions (or hallucinations), which leads to more review, more agent revisions, more tokens burned, and more disconnection from what is being created."
Business Risks Beyond Individual Skill Loss
Beyond individual developer impacts, Faye outlines several organizational risks:
- Vendor lock-in dependency on specific AI providers
- Unpredictable and escalating token costs
- Team paralysis during service outages (Faye notes: "During a Claude outage, numerous LinkedIn posts highlighted that certain developers and engineering teams were at a standstill")
- Atrophied debugging capabilities creating long-term technical debt
Unlike previous programming advances such as FORTRAN, Java, or higher-level languages, agentic coding represents a decrease in deterministic control. Probabilistic systems cannot replace deterministic ones without introducing fundamental ambiguity into the development process.
Recommended Approach: AI as Secondary Tool
Rather than abandoning AI tools entirely, Faye advocates repositioning AI as a secondary tool rather than the primary development method. This approach maintains active coding engagement while delegating selectively to preserve skill development. The critical tension, as Faye emphasizes, is that "success with agentic coding hinges on a skilled developer who's thinking critically and comfortable operating at the architectural level to spot issues in thousands of lines of generated code, yet it's the individual's critical thinking skills and cognitive clarity that AI tooling has been proven to impact negatively."
The article appears amid growing discussion about agentic coding trade-offs, with Anthropic's 2026 Agentic Coding Trends Report and multiple blog posts examining the same tensions between productivity gains and skill preservation.
Key Takeaways
- Developers face a paradox: they need strong coding skills to supervise AI agents, but using those agents causes skill atrophy
- Research documents a 17% decline in skill mastery among developers heavily dependent on AI tools, with debugging capabilities showing the steepest decline
- Agentic coding creates business risks including vendor lock-in, unpredictable token costs, and team paralysis during AI service outages
- Unlike previous programming abstractions, agentic coding represents a decrease in deterministic control by introducing probabilistic decision-making
- Experts recommend using AI as a secondary tool rather than the primary development method to preserve critical thinking skills