The Rise of 'Vibe Coding': Why Bluesky Users Blame AI for Every Glitch

Bluesky recently experienced intermittent service disruptions that triggered a firestorm of criticism from its user base. While the platform officially attributed these issues to an 'upstream service provider,' a significant portion of the community immediately pointed to a different culprit: 'vibe coding.' This term, which describes the practice of relying heavily on AI-powered coding assistants, has become the latest boogeyman for tech enthusiasts skeptical of generative AI. This reaction highlights a deep-seated resistance among many tech users toward the integration of AI in software development. For many, the mere mention of AI tools in a company's workflow is enough to assume that the resulting code is sloppy, brittle, or inherently buggy. The Bluesky development team, in particular, has become a lightning rod for this frustration, with users flooding the platform with memes and angry posts accusing developers of being too lazy or unskilled to write their own code. The tensions were exacerbated by transparent admissions from Bluesky leadership. Founder Jay Graber openly stated that the platform is built with AI, noting that engineers and non-engineers alike utilize Claude Code. Furthermore, Technical Advisor Jeromy Johnson, who goes by 'Why' on the platform, famously claimed that in the past two months, Claude has written about 99% of his code. CTO Paul Frazee joined the conversation by stating that he 'vibecodes' just as much. These comments, while intended to show innovation, served to alienate a segment of the user base that values traditional software craftsmanship. Originally, the term 'vibe coding' referred to amateurs using AI to generate functional but poorly understood code. Today, it has evolved into a pejorative label applied to any software project that admits to using AI. The launch of projects like Attie—a tool designed to help users build custom feeds via chatbot—further fueled the fire. Despite assurances from Graber that the goal is to provide user control rather than generate content, the anti-AI sentiment among Bluesky’s user base remains potent, especially given the platform's previous promises regarding AI training data. Engineers argue that the public perception misses the nuance of how these tools are actually used. They maintain that human-led review, red-teaming, and quality assurance processes remain the bedrock of their development cycle. Tools like Claude Code are viewed as productivity multipliers, not replacements for deep architectural knowledge. However, as some users have pointed out, the cost of adopting AI is the loss of the 'benefit of the doubt.' If a service goes down, the community will now reflexively blame AI, regardless of the actual technical cause. Ultimately, 'vibe coding' has become a catch-all scapegoat for any software failure. While it is true that relying blindly on AI can lead to security risks and buggy software, it is equally true that service outages have been a reality of the internet long before LLMs existed. The ongoing debate suggests that for software companies, the challenge is not just technical—it is a struggle to bridge the gap between their internal development reality and the public's perception of AI-generated 'slop.'