draftless editing, ctrl+c ctrl+v and what the Biles II teaches us about snapshots

I’m not even pretending to compete for your attention these days, but I promised to keep you in the loop. First, the gold medal for ‘Fighting the temptation to check streams from the Olympics while coding’ goes to: The Octoneers team 🥇

The Good

• Copy paste test steps. Ctrl+C and Ctrl+V keystrokes (cmd+c cmd+v for mac users) are most natural desktop shortcuts. We brought them to test editing. Now, you can copy paste test steps within the same test or from other test cases. 👣
• ‍Draftless editing. In order to reduce clicks we ditched the draft state of a test case. Just open the test case and save your changes. We’ll warn you if you try to get out without saving. 💾  ‍
• AI agent benchmark fine tune. Our first implementation of automated benchmarks was a bit flaky. We closed the loopholes and now it works like a well oiled machine. 🛢️

‍

The Biles II of Test Editing

I don’t say it’s an impossible jump, but we still haven’t managed to nail SNAPSHOTS. The virtual locator picker is the most important device for editing our tests. It relies heavily on the displayed snapshot, which must be relevant to the state you want to interact with. Even if you drag the step to a new spot in the sequence.

As Simone Biles shows, as long as it’s within the bounds of physics a lot can be achieved. Getting snapshots right is possible and we are working on it.

‍

The Complicated

The Angular framework’s js-runtime patching caused a slight detour for our dev team. It got into a scuffle with our own patching, which we need to make the page understandable for the AI, causing some older Angular sites to crash for our agent. We moved one level up not to interfere with the framework.

Big thanks to WTF framework issue #555 which pointed us to the right spot in the hierarchy for us to patch.

‍

The Highlight

The AI agent progress indicator bar is back! Going back to a feature you’ve once ditched is so startup. It was discontinued to make room for new features in the app.

It was dearly missed by our users. People being OK with the black-box nature of AI depends on its application. In testing, transparency builds trust. We rebooted the progress indicator in all main views and added it to individual step generation.

At any given moment, you'll have an idea what the agent is doing. No more staring at a spinning circle. ⭕️

Can you spot it?

‍

---

I’d love to hear your feedback on this update. Or the app. Or just say hi.

‍