Prominent Dev Voice Slams AI Code Tools: 'They All Suck,' Citing Early 'Vibe Coding' as Root Cause

A prominent figure in the developer community has ignited debate by declaring that many leading AI-powered development tools, including Cursor, Claude Code, and Codeex, are fundamentally flawed, asserting ‘they all suck.’ This critical analysis attributes the pervasive issues to these tools being ‘vibecoded’ — developed using early, less capable AI models. While dogfooding (using one’s own products internally) is generally considered beneficial, applying this principle prematurely with nascent AI led to significant technical debt. Specific criticisms target Cursor’s inconsistent UI, accidental email leaks, and arbitrary feature deprecations, alongside Claude Code’s non-deterministic behavior, performance bottlenecks, and a user experience described as a ‘slopfest.’ This early reliance on models like Sonnet 3.7 for critical development paths established a foundation of low-quality code, which subsequent improvements by more advanced models struggle to remediate.

This problem is exacerbated by ‘codebase inertia,’ a phenomenon where a project’s quality peaks within 3-6 months, after which bad patterns tend to multiply exponentially, particularly when AI agents accelerate code generation. Current models, while powerful, are more adept at creating new code than cleaning up existing ‘slop.’ To counter this, a radical shift in development discipline is advocated: optimize for ease, clarity, and speed; ‘tolerate nothing’ by aggressively eliminating bad patterns; and embrace ‘sledgehammer development,’ using agents to rewrite substantial portions of a codebase. Furthermore, it’s suggested that engineers must engage deeply in ‘plan mode’ with AI agents, scrutinizing their proposals, and aggressively discarding suboptimal code. A provocative long-term solution involves maintaining parallel codebases: a ‘slopware’ version for rapid prototyping and idea vetting, and a separate, meticulously engineered version for production, mirroring the development strategy seen in games like Vampire Survivors. This approach leverages AI for rapid experimentation while preserving the integrity of production systems.