Blog

How deployment failures hide in pull requests, and how we build Proofline to find them. RSS

  1. From 18 minutes to 8: two days of making our CI fast and cheap

    In two days, the median time to check a merge group on Proofline's main branch fell from 18.4 minutes to 8.0, while the queue handled two to four times the load. The machines running those checks became less than half as expensive per hour. Discarded computation, which consumed half the merge queue's job time in the worst hour we measured, is mostly gone too. Those are three separate gains: shorter checks, cheaper machine-hours and less wasted work.

  2. Week of November 9, 2026

    This week's releases give you earlier review results, more ways to investigate production symptoms, and clearer control over setup and spending. Findings can reach GitHub while the rest of a review is still running. Sentry alerts can start investigations. Review costs are easier to understand, setup offers more direct repair paths, and coding agents can prepare configuration changes for an administrator to review.

  3. Why managing your own AI code review is the wrong investment of your time

    Before asking an engineer to build the reviewer your whole company relies on, decide how much ongoing review work you want your team to own. Your engineers have the judgment to assess a finding. The investment question is how much time they should spend building software that carries that judgment from one review to the next. Most of the work you’re signing up for happens after the first successful review, on the next hundred.

  4. Week of November 9, 2026: faster reviews and more answers from your code

    This week, Proofline got faster and expanded what you can do with a review. Walk through a change, ask about specific lines, and check its impact across repositories. Beyond reviews, you can follow production investigations through to proposed fixes.

  5. Week of November 2, 2026: checks, CI failures, and the work after a finding

    Define your team's pre-merge checks, inspect failed CI jobs beside a review, and follow a coding agent's attempts to fix a finding. This week's releases also add security reporting and more ways to work with Proofline from your editor. The ChatGPT plugin is on its way.

  6. Week of October 26, 2026: finishing recipes, pull request title checks, and Fix all in development

    This release introduces the Fix all workflow, finishing recipes, and a review walkthrough. Built-in fix generation awaits required analyzer checks. It also adds pull request title checks, shadow mode, research, directory sync for single sign-on, and browser-based AWS setup.

  7. Week of October 19, 2026: inspect reviewed changes, ask questions, and fix findings

    Read the reviewed diff beside its findings, ask questions from recorded evidence, and verify fixes from the CLI or a coding agent. This release also adds a codebase wiki, implementation plans, environment architecture documents, a pull request inbox, and Vercel project connections.

  8. Give a review the context it needs

    Run a deeper investigation for a migration, bring the relevant design document into a review, or record which repositories change together. These features are now in the Proofline console, alongside an audit-log export.

  9. Proofline is European

    Built in Europe for teams who want code reviews that account for the infrastructure their code will run on.

  10. AI code review for AWS

    With AI code review for AWS, Proofline checks each pull request against the account it deploys to. It reads supported service, database, network and IAM configuration through a read-only role, then explains how the change could fail there.

  11. BYOK AI code review

    BYOK AI code review runs on your own model provider account. In Proofline, you choose the review models and pay your provider directly for model usage, without a Proofline subscription. Verification VM time, when used, is charged separately in Proofline credits.

  12. Cloud Run deployment review

    Proofline checks how the code changes in a pull request could affect your Cloud Run service. The review uses configuration recorded from Cloud Run, including environment variable names, how traffic is split between revisions, and scaling limits.

  13. Code review audit trail

    Proofline reports potential problems during code review and preserves the findings and supporting evidence. You can return to that record later to see what was flagged and how it was handled.

  14. Full control over your code review

    Proofline settings decide when a review runs, what it reads, what it may block, and what it costs. The audit log records every settings change.

  15. Customizable AI code review rules

    In Proofline, you can write review rules that tell the reviewer what to check across your organization or for a specific environment or repository. Set shared defaults for your organization, then adjust them for individual repositories where overrides are allowed.

  16. Deployment gates

    A merge gate holds a pull request, and a deployment gate holds a deployment, until its review is complete and no blocking finding is open. A fixed policy that you set makes each decision, never the model.

  17. Deployment risk analysis

    Proofline analyzes the deployment risk of every pull request across 18 risk domains. It records which domains the change touched, which it skipped, and why.

  18. Free AI code review for one month

    New organizations get one calendar month of hosted Proofline code review for GitHub pull requests. Private and public repositories are included. You do not need a credit card, your own model key, or a cloud connection to start.

  19. GitHub Actions workflow review

    Proofline reviews GitHub Actions workflow changes under .github/workflows/ for their build and deployment effects, and checks that every external action is pinned. A workflow file decides what ships and where.

  20. AI code review for Google Cloud

    Proofline runs AI code review against the Google Cloud project each pull request deploys to. It reads the project's configuration without a stored Google key.

  21. Low-noise AI code review

    Low-noise AI code review means Proofline publishes a finding only when evidence supports it, says it once, and records the rest without posting. A reviewer that comments on everything trains people to ignore it.

  22. Reviewing AI-generated code

    AI-generated code can look correct without accounting for the environment where it will run. Proofline checks how those changes could affect your production environment. You can run a review from your terminal or on a pull request.

  23. Secure AI code review

    Proofline limits what its AI code review model can reach, so a manipulated model cannot widen its own access. The reviewer reads code from people you do not control, and text that may try to steer it.

  24. Terraform pull request review

    Proofline reviews Terraform and OpenTofu changes at the exact commit under review, and confirms each apply from the GitHub Actions run that performs it. A review can compare a declaration with the configuration Kian's cloud inventory records for the affected resources.