devDosvid blog

An AI agent upgraded a Helm chart, the pipeline said green, and QA was broken for three days. The first instinct was to lock everything down. The better answer was harder to accept.

What harness engineering looks like when the work is infrastructure, not application code.

Most listening advice says add behaviors. The real fix is the opposite — stop doing everything else.

How parallelizing by step instead of by service helped a half-sized team do twice the work

When a teammate proposed a better design than mine, my first instinct was to defend my own — not because of logic, but because it was his idea, not mine.

As senior engineers, we often hit the ceiling at some point: our individual output is high, but writing more code or spinning up one more cluster does not feel like the best way to make more impact. A more significant impact comes from multiplying our efforts. But how do you do that without a team reporting to you? How do you effectively involve peers in your projects, influencing their priorities when you don’t control their backlog or performance review? ...

Legacy setups can hide costs, and sometimes big. By challenging this, I learned some cool stuff about S3 Intelligent-Tiering and Lifecycle Configuration.

Explore AWS outage case studies, uncovering essential strategies for building resilient systems by understanding dependencies and preventing cascading failures

Side-by-side comparison of 6 Terraform static analysis tools: Checkov, Trivy, KICS, Terrascan, tfsec, and Semgrep.

Step-by-step guide to building a serverless backend with API Gateway V2 HTTP API and Lambda using Terraform.