January 2026 Summaries

A discussion of how modern CI systems could improve on GitHub Actions argues for a more flexible, task-oriented design centered on local execution, graph-based parallelism, automatic content-addressed caching, and containerized, portable workflows. It identifies GitHub Actions limitations including slow commit-and-push feedback loops, VM-centric jobs with duplicated setup, manual and potentially unreliable cache configuration, limited remote and local debugging, inefficient retry behavior, oversized base images, proprietary third-party Actions, mutable dependency references, fragmented failure reporting, and static pipeline definitions. Proposed alternatives include starting runs through a local CLI, automatically parallelizing dependency graphs, resuming failed steps rather than entire jobs, using lockfiles for dependencies, aggregating and parsing test failures, retrying individual flaky tests, relying on standard tools such as Docker Compose, assigning resources per task, and generating tasks dynamically at runtime. The post concludes by presenting RWX as a platform that it claims provides these capabilities.

RWX CLI 3.0.0 aims to speed up developer and coding-agent workflows by enabling CI validation before changes are committed or pushed, avoiding the delays of traditional segmented CI processes. New commands allow users to start and wait for runs with `rwx run --wait`, inspect outcomes through `rwx results`, and troubleshoot failures using `rwx logs` and `rwx artifacts`, while all commands support machine-readable JSON output. Leveraging DAG-based execution, content-based caching, sandboxed tasks, and explicit dependencies, RWX lets developers and agents validate changes remotely without requiring local secrets, current development environments, Git pushes, or container builds.

RWX has added log retrieval to its CLI beginning with version 2.3.0, allowing users to run `rwx logs <taskId>` to download task logs for deeper investigation of stack traces and debugging output beyond the platform’s semantic test-result UI. The command supports options such as JSON output and automatic extraction, enabling automated workflows that can pipe log paths into tools like jq and grep to identify errors. Downloaded logs can also be placed where AI agents can access them, while RWX’s MCP server can direct agents to retrieve logs when needed. The approach reflects RWX’s effort to avoid overloading AI prompts with large volumes of log data and to limit MCP tool context, positioning the CLI as a complementary, human- and automation-friendly mechanism for accessing detailed diagnostic information.

RWX has expanded its CI/CD base image support from two predefined Ubuntu options, 22.04 and 24.04, to direct references to any Docker container image using standard image:tag syntax. The former RWX-specific Ubuntu customizations, including common packages, quality-of-life settings, and Docker support, are now provided through the open-source rwx/base package, which currently supports Ubuntu 22.04 and 24.04, while users can select config: none to run other images such as Debian without those additions. The change accompanies RWX’s ability to build container images directly in its runtime and required removing fixed assumptions about shells, users, and workspaces. RWX also automatically produces a container image for every pipeline step, which users can retrieve with rwx pull, leveraging a new container runtime designed to avoid Docker-related performance limitations.