The case against LLMs in software development

Software industry veteran offers a critical analysis of Large Language Models and the degradation of software quality over time.

The core argument: The rise of Large Language Models (LLMs) represents a concerning shift in computing, where corporations prioritize profit over software quality and user experience.

Historical context: Earlier software development emphasized different priorities and characteristics compared to today’s landscape:

• Programs were faster and more efficient despite limited hardware capabilities
• Quality control was paramount due to the difficulty of distributing patches
• Software was typically standalone, purchasable, and didn’t require internet connectivity
• Applications were simpler, focused on specific use cases, and supported multiple hardware architectures
• Independent developers could create meaningful applications with limited resources

Current state of technology: Major tech companies have invested heavily in AI and LLM technology, creating significant market pressure:

• Every major tech company, including Google, Microsoft, Amazon, NVIDIA, AMD, and Apple, has positioned themselves as “AI-first” by the end of 2024
• Traditional services like Google Search and Windows have deteriorated in quality while companies focus on AI integration
• The substantial financial investments in LLM technology make it unlikely that companies will abandon this direction, regardless of effectiveness

Technical concerns about LLMs: Several fundamental issues exist with current LLM implementation:

• Systems are slow, expensive, and non-deterministic by design
• Results can be inconsistent and unreliable
• The technology represents an abstraction layer between users and computing functions
• Environmental impact of massive data center requirements raises sustainability concerns

Generational impact: The normalization of poor software quality could have lasting effects:

• Younger users who have never experienced better alternatives may accept subpar performance as normal
• Growing dependency on LLM tools could reduce understanding of fundamental computing concepts
• Future generations might never experience deterministic software behavior

Industry implications: The focus on LLMs reflects broader changes in the technology sector:

• Competition has diminished across operating systems, search engines, and mobile platforms
• Investment returns often trump technical merit in determining which technologies succeed
• Companies are positioning LLMs as intermediaries for all computer interactions

Looking ahead: The momentum behind LLM technology, combined with massive corporate investments, suggests this trend will continue despite technical limitations:

• The technology sector’s consolidation means few alternatives exist
• Environmental and computational costs may continue to rise
• Individual resistance options are limited to personal choice in software usage and development practices

Market realities and resistance: While the trajectory seems set due to massive financial investments, individual choices remain important:

• Some developers and users are actively choosing to avoid LLM integration in their work
• Small-scale resistance through personal software choices continues
• The future impact of this technology on computing, the economy, and the environment remains uncertain