← Retour au blog
tech 18 July 2026

Faulty Towers, Vibe Sickness, and the Vibe Bobsled: Understanding AI Challenges

The article examines how vibecoded systems pile up code until they become incomprehensible to humans, and how LLMs step in to explain these complex systems, while posing essential questions about the future of software development.

Article inspired by the original source
Faulty Towers, vibe sickness, and the vibe bobsled ↗ dustycloud.org

Introduction

In the ever-evolving world of technology, the rise of vibecoded systems and language models like LLMs (Large Language Models) has sparked intense debate. These systems, often criticized for creating incomprehensible code towers, raise the question: how far can we go in automating software development?

The Tower That Leans

Armin's article "The Tower Keeps Rising" on lobste.rs highlights a concerning trend. Vibecoded systems continue to pile code upon code. The abstractions accumulate, and while these systems may continue to operate, no human can fully understand the codebase. Yet, thanks to LLMs, parts of this code can be explained, allowing development to continue.

Implications

  1. Acceptance of the Status Quo: Advocates of these systems seem to accept this state as the way forward, even if it means developers gradually lose direct understanding of the code.
  1. Pivot to a New Norm: Experts like Simon Willison clearly differentiate "vibe coding" from agentic engineering, insisting that one should be able to explain AI-generated code before integrating it into production.

Vibe Sickness

"Vibe sickness" refers to the phenomenon where developers rely too heavily on LLMs, losing sight of the fundamentals of software development. This creates a dangerous dependency, where human expertise is replaced by machine-generated explanations.

Consequences

  • Reduction in Code Quality: Without thorough human review, the quality of the code risks declining, leading to bugs and inefficiencies.
  • Loss of Skill: Developers may lose their ability to solve complex problems independently, relying too much on automated tools.

The Vibe Bobsled

The "vibe bobsled" is a metaphor describing the speed at which these changes are occurring. In less than a year, the shift from traditional engineering to LLM-guided engineering has been swift.

Concrete Example

Take the example of Claude Code, an LLM agent in software development. When tasked with building a JSON API, without meticulous review, the produced code might be performant but incomprehensible to future developers.

The Future of Software Development

The crucial question is how to balance the use of LLMs while maintaining deep human understanding. This requires a mindset shift and potentially stricter code review policies.

Towards Human-Machine Collaboration

The future may lie in a close collaboration, where LLMs are used to augment human intelligence, not replace it.

Conclusion

Vibecoded systems and LLMs bring undeniable benefits, but they also pose significant challenges. As a decision-maker or developer, it's crucial to navigate this rapidly evolving landscape carefully.

Let's discuss your project in 15 minutes.

Vibecoding LLM AI Software Development Automation
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call