What AI Won't Fix (www.blakeyoder.com)

AI isn't a magic cure for bad engineering habits — it just changes who gets blamed. The piece argues that sloppy code existed long before tools like Cursor or Claude Code and will persist afterward; the difference is that AI becomes the scapegoat. In fast-paced environments where time-to-ship is hours, developers make deliberate trade-offs and accrue technical debt. AI acts as an "intelligence multiplier": used well it accelerates learning (explain DNS, or request flow on ECS/Fargate), but used lazily—pasting a stack trace into a model and accepting the fix—lets people skip the mental work and perpetuate mediocre practices. For the AI/ML community this matters because models amplify both capability and risk. Agentic tools expand the surface area for distraction and side-quests, so guardrails, clear expectations, hard deadlines, code review, tests, and a strict definition of done are more important than ever. The technical takeaway: code quality is not binary or intrinsic to the tool; it depends on human intention, workflows, and constraints. Integrate AI to teach and speed engineers, not to short-circuit learning—use prompts to build mental models, demand explanations for fixes, and maintain discipline to stop when "good enough" is reached. AI will raise the stakes, but it doesn't change the underlying engineering truths.