Imbue releases Offload to split Playwright runs across 200 Modal sandboxes

• Imbue open-sourced Offload, a Rust CLI that uses 200+ Modal sandboxes to spread test runs across remote workers instead of tying up a local machine, according to Imbue's launch thread.
• The launch example is aimed at browser-heavy verification: Imbue says its own Playwright suite dropped from 12 minutes to 2 for about $0.08 per run, while Modal's post describes the result as a 6x reduction in execution time.
• Offload is positioned as a drop-in runner for pytest, cargo-nextest, and vitest with "one TOML config," and the GitHub repo says it can target local processes, shell commands, or Modal-backed sandboxes.

Offload is a new open-source test runner from Imbue built for teams whose bottleneck is verification rather than generation. In the launch thread, Imbue describes it as a Rust CLI that "spreads your test suite across 200+ Modal sandboxes," and the attached Launch demo shows the basic flow: run the CLI, fan work out, and finish with a short config file.

The first target is test suites that are expensive to parallelize locally, especially browser automation. Imbue's announcement uses a Playwright run as the headline benchmark, cutting wall-clock time from 12 minutes to 2 at $0.08 per run, while Modal's summary frames the same result as giving agents a way to parallelize tasks for a 6x speedup.

The repository fills in the implementation details missing from the social posts. According to the repo summary, Offload supports pluggable execution providers for local execution, shell-based runners, and Modal cloud environments, which makes it broader than a Modal-only wrapper. The same source says it already works with pytest and cargo test tooling, includes JUnit XML reporting, and can retry flaky tests automatically.

Imbue's repo link also suggests the scheduler is doing more than naive sharding. The project description mentions "intelligent scheduling based on historical timing data," plus group-level filtering and environment-variable expansion. That makes the tool relevant for CI stacks where test duration is uneven and browser jobs dominate the critical path, not just for agent demos that need more parallelism.