
ハーネスエンジニアリングを極めたら、IssueからAIエージェントが動き、人間の役割は要件定義だけになった
https://zenn.dev/aicon_kato/articles/harness-engineering-startupThe evidence pack includes HTML, screenshots, summaries, and metadata. It can be downloaded on Pro.
ハーネスエンジニアリングを極めたら、IssueからAIエージェントが動き、人間の役割は要件定義だけになった
Open the archived HTML with saved-time metadata attached.
This HTML has CSS and images embedded, so it can still be opened even if the original page disappears.
This page describes the implementation of harness engineering in a startup environment. With limited full-time engineers, the team built an autonomous development pipeline with 21 AI agents over two months. February focused entirely on CI/CD improvements, while March alternated between feature development and pipeline optimization. Using GitHub Issues-driven automation and Claude API, the pipeline automatically generates PRs ready for merging from initial requirements. By late March, 681 PRs were merged bi-weekly. The system leverages GitHub Actions, Claude API, and Terraform to automate all development processes, leaving only requirement definition to humans.
