Don't Try This at Home: What Happened When I Sent Kiro to Work and Went Gardening
There is a special kind of confidence that comes from an AI development environment that just works. You describe a feature, and it writes the Terraform, the Go, the Lambda handlers, the S3 wiring. It deploys. The happy path works. And you think: this is fine. I can go do something else now. So one afternoon I handed Kiro a task, hit go, and left it running. Right on cue, I was handed a task of…
One afternoon, I entrusted an AI named Kiro with a serverless recipe service project on AWS. The system automatically generated Terraform, Go code, Lambda handlers, and S3 wiring. I left Kiro to work while I attended to a gardening task. Later, I received an alert from AWS stating that a recursive loop had formed between a Lambda function and an S3 bucket, causing the function's invocations to be disabled.
The recipe service allowed users to upload photos and automatically generated thumbnail and display versions of each image. My instruction to Kiro was to generate the two derivative versions from each uploaded image, without specifying where they should be stored. Kiro created a Lambda function that triggered when a new object was uploaded to the S3 bucket, resized the photo, and saved the thumbnails and display versions back into the same bucket.
This created a recursive loop as the Lambda function kept triggering and writing new objects into the bucket, causing it to process images indefinitely. The issue is a well-known S3 footgun, which warns against triggering a Lambda function from an S3 bucket and writing back to the same bucket without a guard. To fix this, I used a naming convention with a unique suffix for the original photos and scoped the S3 notification to only trigger on the original suffix, effectively breaking the loop.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
