I will not be using LLMs in my coding
Source Entity
Hacker News
A developer has publicly declared a decision to abandon Large Language Models (LLMs) in their coding workflow. This shift highlights growing industry debates regarding the efficacy, security, and long-term impact of AI-assisted programming tools.
The Shift Away from AI-Assisted Development
Recent discourse within the developer community has highlighted a growing trend of professionals choosing to move away from Large Language Models (LLMs) in their daily coding workflows. This decision reflects a critical reassessment of how AI tools, originally intended to boost productivity, may impact the fundamental craftsmanship of software engineering.
Concerns Over Code Integrity and Dependency
One of the primary motivations for this departure is the concern over code quality. While LLMs excel at generating boilerplate text and common functions, they frequently produce non-deterministic outputs that can introduce subtle bugs or security vulnerabilities. Developers opting out are prioritizing the deep understanding required to write, debug, and maintain code manually, arguing that reliance on AI-generated snippets leads to a degradation of essential problem-solving skills.
Security and Intellectual Property Risks
Beyond technical proficiency, the decision to stop using LLMs often stems from heightened awareness of security and intellectual property risks. Organizations are increasingly cautious about feeding proprietary codebases into external AI models, fearing potential data leaks or the accidental inclusion of copyrighted code. For many, the risk of violating licensing agreements or exposing sensitive logic outweighs the convenience of AI-powered autocomplete features.
The Impact on Developer Skill Acquisition
There is a pedagogical argument that using LLMs as a crutch hinders the development of junior engineers. By skipping the "struggle phase" of debugging and algorithm design, new developers may fail to build the mental models necessary to navigate complex system architectures. Those eschewing AI tools are often seeking to reclaim this foundational learning process, ensuring that they understand the 'why' behind the code, not just the 'how.'
Future Trends in Software Engineering
This movement suggests a potential bifurcation in the industry: one camp leaning heavily into AI-driven automation for speed, and another prioritizing human-centric, verifiable, and highly secure code. As the novelty of AI coding assistants wears off, the industry is likely to move toward a more balanced approach, where developers integrate AI tools only for specific, low-risk tasks while maintaining manual control over critical system components.
Conclusion
The declaration to stop using LLMs in coding is more than a personal preference; it is a signal of a broader professional maturation regarding generative AI. As developers weigh the efficiency gains against the risks of technical debt and skill attrition, the industry will likely see a push for more transparent, secure, and developer-controlled AI implementations that respect the integrity of the software development lifecycle.