← back
Automating Large Scale Refactors with Parallel Agents - Robert Brennan, OpenHands
Takeaway
Parallel agent orchestration in cloud sandboxes is unlocking automated remediation of tech debt at a scale impossible for single-agent or human workflows.
Summary
- Robert Brennan (OpenHands co-founder/CEO) traces coding-agent evolution: context-unaware snippets, context-aware autocomplete (Copilot), autonomous agents (Devin/OpenDevin/OpenHands in early 2024) and now parallel orchestrated agents.
- Frames the adoption curve: median devs use local agents like Claude Code; early adopters use cloud-sandboxed agents that can run in parallel without rm-rf risk; top 1% are orchestrating fleets where agents spin up sub-agents.
- OpenHands (MIT-licensed, formerly OpenDevin) ships both a cloud platform and a local CLI to meet developers where they are.
- Best orchestration use cases are repeatable: dependency bumps, CVE remediation, large-scale refactors, modernization — one client uses OpenHands to remediate CVEs across many repos.
- Going from IC coding to dispatching agent fleets feels like the IC-to-manager transition; community-driven open source matters because software dev workflows shouldn't be a black box.
agentscode-generationopenhands
Original description
Today's agents are best at small, atomic coding tasks. Much larger tasks--like major refactors and breaking dependency updates--are highly automatable but hard to one-shot. In this session, we'll discuss patterns for orchestrating large-scale code changes with swarms of agents and a human in the loop. We'll also work through a concrete example: migrating an entire codebase from one React state management library to another. https://twitter.com/RobertBrennan Slides: https://dub.sh/openhands-workshop