Software Engineers Say They’re Losing the Ability to Code Now That AI Does It for Them

• Engineers report forgetting fundamental skills like API implementation after months relying on AI code generation instead of writing themselves.
• Meta and other major employers are tracking AI token usage via leaderboards, creating implicit pressure for engineers to adopt code-generation tools.
• Developers warn that outsourcing thinking to AI degrades critical reasoning and design skills needed for complex architectural work.

• 404 Media interviewed multiple developers confirming skill erosion complaints proliferating across social media platforms.
• Meta introduced real-time AI token-usage leaderboards to measure adoption among engineering staff.
• Cursor widely distributed among engineers who then adopted it regularly, illustrating seductive nature of tools regardless of formal mandate.

A significant skill-degradation problem is emerging within engineering teams at major technology corporations, as the widespread adoption of AI code-generation tools is replacing hands-on coding work with passive code review.

Engineers who once wrote software daily now spend their time validating AI-generated code, a rote activity that fails to exercise the problem-solving and architectural reasoning skills required for complex systems.

The shift is not merely a matter of convenience or efficiency; it represents a structural change in how engineering work is performed, with consequences that extend beyond individual productivity to organizational capability.

The scope of the problem is substantial enough that complaints are no longer isolated incidents confined to private conversations. Multiple developers have come forward to describe the specific ways in which reliance on AI tools has eroded their technical competency.

One engineer with years of professional experience and a university degree in computer science reported being unable to implement a Laravel API without AI assistance, a task well within the scope of standard backend engineering work.

I had some issues where I forgot how to implement a Laravel API and it scared the s**t out of me. I went to university for this, I’ve been a software engineer for many years now and it feels like I am back before I ever wrote a single line of code.

Anonymous software engineer

This account illustrates not a temporary lapse but a genuine erosion of core competency.

Meta’s Real-Time Leaderboards Create Adoption Pressure Among Unwilling Engineers

The normalization of AI code generation is not happening organically through voluntary adoption alone. Major employers including Meta have introduced measurement systems designed to track and publicize AI tool usage among their engineering workforce.

Meta’s leaderboards display which employees are consuming the most AI tokens in real-time, creating a visible performance metric that transforms tool adoption from an optional convenience into an implicitly mandated practice.

The leaderboard approach is particularly effective at driving adoption because it combines technical measurement with social signaling. Engineers observe which colleagues rank highest on the AI usage metrics, creating peer pressure and competitive dynamics that encourage increased reliance on automation.

The result is that adoption becomes decoupled from individual choice or assessment of actual productivity gains, instead becoming a marker of alignment with company priorities and efficiency expectations.

Not all companies have formalized this pressure through explicit leaderboards, but the competitive and seductive nature of AI tools drives adoption regardless.

Another developer described a subtler but equally effective pressure mechanism: broad access to AI coding tools like Cursor distributed without formal mandate, but presented as a standard development resource.

In this scenario, engineers found themselves regularly using the tools not because they were required to, but because the tools were ubiquitously available and normalized within the team environment. The distinction matters because it obscures corporate responsibility for adoption choices while producing identical outcomes, engineers spending less time writing and more time reviewing.

Outsourced Thinking Undermines Critical Problem-Solving Across Design and Architecture

The risk extends beyond simple skill loss to encompass degradation of the high-level reasoning and design capabilities that distinguish senior engineers from junior ones. As one developer explained, the problem is not merely about forgetting syntax or API structure, it is about the outsourcing of thought itself.

When engineers reflexively turn to AI to generate answers rather than reasoning through problems independently, they atrophy the capacity to identify potential issues, evaluate multiple architectural approaches, or defend design decisions.

This developer characterized the experience as mental degradation comparable to the loss of phone number memorization after smartphones became ubiquitous, but operating at a deeper and more consequential level.

Where forgotten phone numbers represent a loss of information retrieval, forgotten problem-solving and design reasoning represent a loss of intellectual capability necessary for senior technical work. The engineer noted explicitly that “critical thinking and ability to sit and reason about a problem or a design has degraded because the all-knowing-dalai-llama is just a question away.”

The concern is not theoretical but practical: employers who systematize the outsourcing of engineering thought may find themselves unable to execute complex, novel, or high-stakes technical projects when they require deep reasoning rather than code generation.

Accumulating research evidence supports these firsthand accounts. Studies examining the impact of AI assistance on cognitive function show patterns consistent with skill atrophy across technical and non-technical domains.

As AI systems become more capable and more integrated into professional workflows, the conditions that preserve and develop critical thinking skills, struggle, iteration, error recovery, and reasoning through novel problems, become less frequent and less valuable within the work environment.

Long-Term Workforce Capability Risk for Enterprises Betting on Automation Scale

The implications extend directly to enterprise risk and the sustainability of productivity claims made by corporations investing heavily in AI automation infrastructure.

Companies that systematize AI tool adoption without preserving opportunities for hands-on technical work face a medium-term problem: an engineering workforce that has lost the foundational skills required to handle complex problems that AI cannot solve, maintain legacy systems, or pivot to new technical domains.

For institutional investors and corporate buyers of AI infrastructure, the finding suggests that the productivity gains celebrated by software companies may be partially illusory, representing not genuine efficiency improvements but rather a substitution of high-skill technical work with lower-skill code review and validation.

The efficiency is real in the short term, but the long-term cost is reduced organizational capability and increased dependency on external AI systems for any novel or complex work.

The problem is not uniform across all employers or all engineering disciplines.

Companies that have not formalized AI adoption or that actively preserve hands-on coding work may maintain both productivity gains and workforce capability. Engineers in specialized domains like systems programming, security research, or infrastructure architecture may have less opportunity to outsource their thinking because those domains require continuous hands-on problem-solving.

The question for institutional stakeholders is whether the corporations most aggressively pursuing AI automation, and publishing the largest productivity gains, are the same ones experiencing the most severe skill atrophy among their engineering teams, and whether those gains will persist as the cognitive debt compounds.

The resolution of this tension will likely emerge over the next 18 to 36 months as major technology companies face complex technical challenges, security incidents, architectural failures, or novel system requirements, that require senior engineering judgment rather than code generation. Whether those companies can rely on their current engineering workforce to address such challenges, or whether they must rebuild internal technical capability or hire external expertise, will determine whether the AI productivity gains hold or represent a temporary efficiency improvement followed by increased technical debt and capability loss.

Original reporting: futurism.com