agents understand their own success better now

I skipped a few updates because of the our launch, sorry for that. Let’s get to what we shipped, what we struggled with and what we nailed last month without further delays.

The Good

AI agents

• We have rethought our page purpose agent. It’s a specialized agent identifying the intent of a web app (see last update). Its more detailed version is an integral part of every AI generation now, not just the initial test discovery. Even if you don’t see it in the UI now, it's looped into every AI run. 🤖
• We’ve improved handling of radio buttons, invisible elements, and dynamic content - e.g. complex dropdown selection in forms. 🔘
• Agents have now a better understanding of their own success. This helps them improve their own output. ✅

Test runner

• You’ll get to your first test report quicker when starting a new project. ⏩
• More robust test execution to overcome hydration issues of test targets 🫗

UX 

• Test reports have a new home now. On the test report overview page you can oversee all test results, launch new test runs and schedule to run them regularly. 📊
• We added a short survey to the sign-up to figure out what your testing priorities are. We prefer asking you directly to intrusive tracking. 📋
• We show the run status of the test case in dependency. So you see what part of the chain didn’t work right away. ⛓️
  

Editing

• We have a multi-select option in test steps, now. With hotkeys. 🔥
• We added test step deletion hotkeys - cmd/ctrl for selecting MULTIPLE single elements, and shift for selecting a RANGE. 🗑️

System

• If you want to test a web app that is not publicly available, no problem! We built a private location worker to help you with it. 🔒
• We introduced a new pro plan - no limit on projects, tests and use of our AI features. Easy to upgrade to from the app now. Compare our plans here. 🏷️

‍

The Bad

Chrome caught a seasonal flu. We encountered several unfortunate "Aw, Snap!" loading errors. We are on it.

‍

The Complicated

Migrating to eslint v9. We usually appreciate eslint for everything they are doing, but this was NOT a fun couple of days. 29+ page migration guide, migration command not broadly supported, issues to set up a config that works. Getting it to typecheck was another treat, best described by our lead engineer: “When you are at the point to run all the linting with type checking - you get 25.000 issues in your monorepo that need a fix and produce a gajillion merge conflicts for everyone."

If you had a similar experience, drop me a note. We have devs who can bond over shared trauma.

‍

The Highlight

Our new dependency tree view is not just to quickly overview your test suite and have the AI agents generate more tests. It’s a visual model for how we structure the quickest path to a higher test coverage of your web app.

You see clearly where to expand the tests following others using AI auto-generation.

‍

---

Btw, thanks a lot for your support with Product Hunt. We landed the Product of Day top rank! New users tested the new build and we got tons of fresh feedback.  

Talking about feedback, this is my calendar where you can book a session with me, I'd like to hear what you think.

‍