Two engineers involved in the OpenClaw project — also called 'Lobster' — have raised the alarm: AI tools may speed up development, but they can also churn out low-quality and potentially dangerous code directly into production. Their concern isn't with AI itself, but with how quickly companies are leaning on these tools without adequate oversight.
Mario Zekner, the creator of OpenClaw's internal agent framework Pi, says the infrastructure is already under severe pressure, and software is getting more vulnerable by the day. The industry might look the other way for a while, he adds, but eventually it will pay for the accumulated mistakes. Fellow engineer Armin Ronacher argues that AI should help experienced developers move faster. Instead, many firms treat it as a shortcut to immediate cost savings, sacrificing long-term quality.
The real problem is that generating code isn't the same as doing proper engineering. AI can quickly spit out drafts, standard functions, and prototypes. But without review, testing, and architectural oversight, that code often creates more problems than it solves. The worst cases happen when generated snippets go straight into live products — skipping security checks, load testing, compatibility analysis, and future maintenance planning.
The engineers warn that chasing short-term productivity could lead to a shortage of junior talent, ballooning technical debt, fresh security holes, and more frequent service failures. When companies replace developer training and growth with automated code generation, they get instant speed today — but tomorrow they lose team stability and product quality.
The Wall Street Journal calls this phenomenon 'vibe slop' — a blend of 'vibe coding,' where development relies on quick prompts and rough results, and a flood of AI-generated garbage. For businesses, the message is clear: these tools can be powerful helpers, but they don't remove the need for solid engineering practices. Critical systems still demand experienced developers, code reviews, refactoring, testing, and security audits.
In the end, the time saved upfront often becomes a bigger expense later. If AI produces low-quality code, the true cost of development is just pushed into the future — into bug fixes, outage investigations, vulnerability patches, and rebuilding user trust.