OpenHands 1.0 dropped - autonomous coding agent now solving 68% of SWE-bench Verified (the harder, human-vetted subset). That's production-grade performance.
Ships with Docker sandboxing for isolation and built-in policy controls so it doesn't wreck your codebase. No more "experimental toy" disclaimers - this is deployable infrastructure.
SWE-bench Verified is the real test: actual bug fixes from real repos, not synthetic problems. 68% pass rate means it's handling non-trivial PRs autonomously.
Open source agentic coding just crossed the threshold from research demo to actual dev tooling you can run in CI/CD.
Ships with Docker sandboxing for isolation and built-in policy controls so it doesn't wreck your codebase. No more "experimental toy" disclaimers - this is deployable infrastructure.
SWE-bench Verified is the real test: actual bug fixes from real repos, not synthetic problems. 68% pass rate means it's handling non-trivial PRs autonomously.
Open source agentic coding just crossed the threshold from research demo to actual dev tooling you can run in CI/CD.