These are 21 selected public repos, not an industry benchmark. The report has more findings and lets you inspect each project separately.
Report + data: https://t.co/MjuNWmhH5S
What's eating most of your E2E maintenance time?
I went through 549,224 commits to understand why E2E tests keep changing. To do it, I built HekaJev on top of Jev and open-sourced it, so you can ask questions of your own Git history too.
I didn't expect what I found.
I think the bigger opportunity for AI is maintenance. It has to work out why a test failed, what changed in the product and whether the check is still worth keeping. That's harder than writing another test. Otherwise new coverage becomes another pile to maintain.
Opus 4.7 is stunning
a refined but refreshing evolution of something familiar
a clear benchmark, and a real sign that a new era has begun
(i haven't tried it yet)
Just open-sourced Qalti. An AI agent that automates any task on your iPhone by tapping the screen like a human. Built for QA but works for any iPhone automation. MIT license. Give it a spin: https://t.co/z7c3R6qj6o
Fun new update: bash access
Now we can run custom commands (send a push via simctl, run bash script to create clean user, or just run subtest with preconditions)
Use https://t.co/kTXWNQl0m1 to automate your manual testing