Don't Try This at Home: What Happened When I Sent Kiro to Work and Went Gardening

A developer recently shared a cautionary tale about using an AI coding assistant to build a serverless image processing pipeline on AWS. While the AI successfully generated the necessary Terraform and Lambda code to resize images, it created a critical architectural flaw: the Lambda function was configured to write processed images back into the same S3 bucket that triggered the function. This resulted in an infinite recursive loop, triggering thousands of unnecessary invocations until AWS automatically disabled the function. The author emphasizes that while AI tools excel at handling tedious plumbing, they lack the architectural judgment to identify feedback loops. The incident highlights the importance of manual oversight, implementing guardrails like prefix filtering or separate buckets, and monitoring cloud metrics immediately after deployment to catch runaway processes before they escalate. The author concludes that while AI significantly accelerates development, the responsibility for system safety remains firmly with the human engineer.