Automating Large Scale Refactors with Parallel Agents
summary
Claims that parallel agents orchestrated from a cloud sandbox can automate massive refactors and dependency fixes at scale, using the Open Hands agent SDK and a refactor SDK to batch files by dependency graph, run verifier/fixer agents in Docker, and open tidy PRs for human review. A corporate user sees about a 30x improvement in CVE resolution time; an actual frontend migration from Redux → Zustand completes component-by-component with scaffolding and human approvals, and the system produces merge-ready PRs.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:11AI coding assistance evolution
- 8:47Agent adoption market
- 13:59Single agents limitations
- 18:14Open Hands Refactor SDK demo
- 27:35Task decomposition strategies
- 32:06Context sharing in parallel agents
- 35:15Live coding setup
- 51:57Debugging agents
- 1:04:27Parallel CVE remediation
- 1:15:05Q&A
related talks
26mOrchestrating AI Agents for Large-Scale Refactoring
17mSoftware Development Agents: What Works and What Doesn't
32mFrom Toil To Triumph: Harnessing Agentic AI To Streamline Infrastructure as Code
23mManaging fleets of coding agents with OpenHands
42mRevisit Old Problems with New Eyes!
51mFuture of Software Development
21mAutomating Front-End Refactoring
36mAgentic CI/CD: from Pull Request to Production without the Paper Cuts