<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>preloop — writing</title><description>preloop is CI built to do less work. Your GitHub Actions workflows run unchanged. Steps and jobs that did not change are reused, failed jobs pause instead of vanishing, and agents can drive the whole loop over one API.</description><link>https://preloop.dev/</link><language>en-gb</language><item><title>GitHub Actions now supports workflow lockfiles</title><link>https://preloop.dev/blog/github-actions-workflow-lockfiles/</link><guid isPermaLink="true">https://preloop.dev/blog/github-actions-workflow-lockfiles/</guid><description>A lockfile turns action tags into recorded commits. That removes a class of supply-chain risk, but only if CI checks it before a job runs.</description><pubDate>Sat, 07 Nov 2026 09:00:00 GMT</pubDate><category>ci</category><category>security</category><author>preloop engineering</author></item><item><title>How preloop decides which tests to skip</title><link>https://preloop.dev/blog/how-preloop-decides-which-tests-to-skip/</link><guid isPermaLink="true">https://preloop.dev/blog/how-preloop-decides-which-tests-to-skip/</guid><description>preloop skips a test only when its detected inputs are complete, unchanged, and safe to trust.</description><pubDate>Wed, 28 Oct 2026 09:00:00 GMT</pubDate><category>testing</category><category>incremental</category><category>correctness</category><author>preloop engineering</author></item><item><title>A fork PR can cancel your production deploy</title><link>https://preloop.dev/blog/fork-pr-cancel-production-deploy/</link><guid isPermaLink="true">https://preloop.dev/blog/fork-pr-cancel-production-deploy/</guid><description>A concurrency group keyed on a branch name lets any fork cancel a deploy. The same shape appears in workflow resolution, environment gates, OIDC and artifacts.</description><pubDate>Sat, 24 Oct 2026 09:00:00 GMT</pubDate><category>security</category><category>ci</category><author>preloop engineering</author></item><item><title>Pause a failed CI job instead of rerunning it</title><link>https://preloop.dev/blog/the-missing-debugger/</link><guid isPermaLink="true">https://preloop.dev/blog/the-missing-debugger/</guid><description>Start a CI job with --debug, and preloop keeps a failed step&apos;s machine and workspace available for debugging and retry.</description><pubDate>Wed, 21 Oct 2026 09:00:00 GMT</pubDate><category>debugging</category><category>ci</category><category>agents</category><author>preloop engineering</author></item><item><title>What preloop caches and how it stays correct</title><link>https://preloop.dev/blog/every-cache-explained/</link><guid isPermaLink="true">https://preloop.dev/blog/every-cache-explained/</guid><description>preloop serves eighteen cache protocols from one content-addressed pool with immutable entries and scoped lookups.</description><pubDate>Wed, 14 Oct 2026 09:00:00 GMT</pubDate><category>cache</category><category>architecture</category><category>correctness</category><author>preloop engineering</author></item><item><title>Why we rewrote the GitHub Actions runner in Rust</title><link>https://preloop.dev/blog/github-runner-in-rust/</link><guid isPermaLink="true">https://preloop.dev/blog/github-runner-in-rust/</guid><description>The official runner carries a .NET runtime and two Node versions into every job. Ours is 36x smaller and idles at a seventh of the memory.</description><pubDate>Sat, 10 Oct 2026 09:00:00 GMT</pubDate><category>runners</category><category>architecture</category><category>ci</category><author>preloop engineering</author></item><item><title>Typed CI with generated YAML</title><link>https://preloop.dev/blog/workflows-as-real-code/</link><guid isPermaLink="true">https://preloop.dev/blog/workflows-as-real-code/</guid><description>Generate provider-compatible workflow YAML from typed source, keep the YAML in the repository, and check for drift in CI.</description><pubDate>Wed, 07 Oct 2026 09:00:00 GMT</pubDate><category>typescript</category><category>ci</category><category>developer-experience</category><author>preloop engineering</author></item><item><title>Deciding which state to explore next in property testing</title><link>https://preloop.dev/blog/property-testing-state-exploration/</link><guid isPermaLink="true">https://preloop.dev/blog/property-testing-state-exploration/</guid><description>Property testing gives you an oracle. The hard part is choosing the next action, and a model-guided explorer found the right one about 3x more often than random.</description><pubDate>Sat, 26 Sep 2026 09:00:00 GMT</pubDate><category>testing</category><category>agents</category><author>preloop engineering</author></item><item><title>The models I daily-drive for CI repair</title><link>https://preloop.dev/blog/models-for-ci-repair/</link><guid isPermaLink="true">https://preloop.dev/blog/models-for-ci-repair/</guid><description>On CI-bench, the models on the cost frontier are also the cheapest. For CI repair, reliability per dollar beats the largest model.</description><pubDate>Sat, 12 Sep 2026 09:00:00 GMT</pubDate><category>ci</category><category>agents</category><category>testing</category><author>preloop engineering</author></item><item><title>Do you really need remote CI?</title><link>https://preloop.dev/blog/do-you-really-need-remote-ci/</link><guid isPermaLink="true">https://preloop.dev/blog/do-you-really-need-remote-ci/</guid><description>Most of what remote CI does is the same work you could run where the change was made. The useful question is which checks still need to happen somewhere else.</description><pubDate>Sat, 29 Aug 2026 09:00:00 GMT</pubDate><category>ci</category><category>agents</category><author>preloop engineering</author></item><item><title>CI-bench: how good are agents at diagnosing and fixing CI failures?</title><link>https://preloop.dev/blog/ci-bench-agents-diagnosing-ci-failures/</link><guid isPermaLink="true">https://preloop.dev/blog/ci-bench-agents-diagnosing-ci-failures/</guid><description>We ran 80 real CI failures through nine agent configurations, three times each. First-attempt scores hid real differences in reliability.</description><pubDate>Sat, 15 Aug 2026 09:00:00 GMT</pubDate><category>ci</category><category>agents</category><category>testing</category><author>preloop engineering</author></item></channel></rss>