Release notes··3 updates

    Code-aware PR reviews powered by the Assessment Agent

    3 updates shipped on April 22, 2026.

    1. Update 01 of 03

      The Assessment Agent now drives PR reviews

      Structured verdicts replace freeform analysis in PR review chats.

      When QA.tech runs tests on a PR, the assessment of each test (passed, failed, why) is now produced by the Assessment Agent and reused throughout the PR review flow. The PR review chat reads the existing assessment instead of running a fresh analysis from scratch.

      In practice this means PR reviews are faster, cheaper, and consistent with what you see when you open the run directly. The verdict, key observations, expected vs actual behavior, and confidence score are the same numbers everywhere. Bot comments now show three states – confirmed failure, needs review, passed – driven by the assessment's confidence tier, instead of a binary pass/fail.

      # Direct link

    2. Update 02 of 03

      Marketing site content routing

      • Added blog, case studies, and changelog sections backed by Markdown in the repository.
      • Wired navigation to real internal routes for faster iteration without a CMS.

      # Direct link

    3. Update 03 of 03

      PR reviews now read your code

      The PR review agent has access to the diff and grounds every suggested test in actual changed code.

      PR reviews can now fetch and reason about the actual diff in your pull request. Every test the agent suggests on a PR must cite a specific file, hunk, or symbol from the change, with a direct citation back to that file. If the agent can't ground a suggestion in the diff, it skips it instead of inventing one. This sharply reduced the rate of "Blocks merge" findings that weren't actually about the change.

      The agent now also fetches the latest deployment for the PR rather than only what the most recent commit produced. If your last commit was a README update or a config tweak that didn't trigger a redeploy, the review still runs against the actual preview.

      When a PR references an issue tracker ticket (Linear, Jira), the agent fetches that ticket automatically so the review is grounded in the requirement, not just the diff.

      # Direct link

    Ready to end the QA bottleneck?

    See how QA.tech agents test your product in a 30-minute demo – and leave with a plan to reclaim those hours.

    Get a demo