How to kill a Zombie?

We all have our software "Zombies" – pet projects and ventures that keep resurfacing stronger each time we try to eliminate them. One such Zombie is Contenox, an AI system designed for optimal performance with non-frontier models and local models. It boasts competitive performance against frontier harnesses, adequate token caching, proper multi-layer security, and a potentially high-margin distributed LLM proxy router.

The allure of Contenox lies in its promising feedback from pilot users, tons of spare cloud credits, and budget for user acquisition. However, the challenge arises when we need to alter minor operational details, such as updating permissions or configurations. The Zombie rises from the dead, claiming more time, resources, and even affecting our health and finances.

The issue stems from AI-generated code that lacks proper architecture and design thinking. When a new requirement is introduced, the Zombie springs back to life, demanding another round of coding and deployment. The code review process fails to catch these bugs, as AI-generated code often leaves no trace for reviewers to identify.

To "kill" a software Zombie like Contenox, we must first acknowledge that relying on AI to write code is not the answer. Software engineers should strive to build their products with human intuition and critical thinking. By avoiding the pitfalls of AI-generated code and embracing a more hands-on approach, we can prevent the creation of these elusive and resource-draining Zombies.