july feature drop // part 3

Ever since agents based on the latest LLMs became advanced enough to act autonomously, questions about how humans interact with them and maintain control of the process have become much more important. We’ve had to figure out:

• How much freedom does the Agent get when discovering test cases? Do our users want to magically get 3 tests or do they prefer being in control every step of the way? Our users are definitely on the “being in control” side
• How much of what the Agent is doing do our users want to see? As much as possible!
• How can you plug in the human into the test generation process and switch back to the Agent? Very different processes for non-coding testers and developers.

In the past weeks, we’ve shipped a lot of features to provide full insight into what the Agent is doing, to enable users to take over or switch back to the Agent via the UI, and to offer test recommendations to choose from - rather than letting the Agent freely auto-generate tests and surprise you with the output.

Introducing controlled test discovery and generation:

‍

Recommended tests to your taste

When you create a new project in Octomind, our AI Agent is going on a discovery journey of your site. It takes a screenshot and tries to come up with a few suggestions for relevant tests to start with automatically.

‍

‍

You can easily select one to have it generated - you’ll be guided through every step the Agent takes. And whenever you’ve created a new test case, Octomind will come up with further proposals you can pick from.

More about recommendations in docs

‍

Timeline to go back and forth in history

And a LIVE timeline displaying what goes on during AI generation

‍

‍

We have recreated the timeline as a centerpiece for our UI to visualize:

• what the Agent has done in your app or is doing at the moment
• What exactly happened every time we have run a test
• What happened in tests executed prior to this one (dependencies)
• Controls to manipulate and edit test steps visually in hi-res snapshots

You will still be able to go through a list of steps, but a visual insight is a more intuitive option for many of our users.

‍

Go to app and play with the timeline

‍

Slider to control the take over from the Agent

An inconspicuous widget in the timeline got some powers

‍

‍

Now, you can take the slider and move it back and forth to:

1. select the exact spot you want to
2. take over from the agent
3. record and edit the test case yourself
4. hand over back to the agent to generate further steps, if needed

It’s a controlled and smooth way to have human users co-create tests with the Agent.

___

‍

Faster QA for early movers in AI sales coaching

A case study.

‍

We are very proud to show you how Anom and Will from SelfActualize.AI use Octomind to automate QA while scaling the development of their sales coaching app.

See how they mastered QA

‍

---

‍

This has been a busy summer so far. Let me know what you think about the new stuff - we rely on your feedback to do this right.

‍

FYI, we are not slowing down any time soon.

‍