Starting today, I plan to send a bi-weekly product update summarizing our work and progress over the past 2 weeks. This will be the only newsletter from Octomind. We want to keep you informed without flooding your inboxes.
The Good
- In-app & email notifications. Unpopular but necessary. You want to be alarmed when your app broke, even if it's nagging on you. We'll now let you know in the app once your test run has finished or our AI agent completed its discovery. We will ONLY send you an email if your test report is red. 🔴
- Test case grouping. Any list of more than 10 tests gets really disorienting. That's why we will now group your test cases. We started with the “link tests” category and will add more soon. 🔗
- Hotkeys for editing test steps help you navigate between steps using arrow keys. ➡️ ⬇️
- Progress indicators. We tweaked several indicators to give you more insight into your test progress without adding complexity. ✅ 🚫
- AI Agent indicators. At any given moment, you'll know what the agent is doing during discovery or when it’s running in drafts. No more staring at a spinning circle ⭕️ wondering when it will finish.
- Status icons for test cases and step results. Done. Check that box complete! ✔️
Check out our improved app 👉 app.octomind.dev
The Bad
Our functionality ambitions sometimes conflict with UX simplicity. We are doing our best to catch ourselves to ensure we are balancing user experience and efficiency with the controls you need to do your job.
- Exhibit A - Buttons
There are too many confusing buttons, and they aren’t always where you expect them to be. We have a list, and are tweaking it, one by one.
- Exhibit B - Names
Names are one of the 2 most difficult things in programming. And boy, even harder in user design. Test discovery, generation, approval, validation, running, saving, publishing…We’re working on improving our naming nomenclature so that our names are clear, understandable, and highly intuitive for all of our users.
The Complicated
We took the first steps towards moving away from LangChain. Although it’s the leading framework for making LLMs work for your AI agents, it has become too complex and rigid and is limiting our ability to innovate quickly.
The Highlight
Since users have told us that AI-based Test Discovery is one of their favorite features, that’s a place where we are doubling down. AI agent enhancements we shipped in the last 2 weeks include:
- Improved world model. This significantly improves our ability to understand context and derive the correct actions.
- Fixed duplicate navigation. The agent now performs better and reduces the frequency in which agents get stuck in loops.
- SVG. We can now handle SVG elements like images and icons.
- Pop-up modal. Octomind can automatically deal with pop-up modals without issue.
Thanks for reading. I look forward to hearing your suggestions on what you want to know more about and how we can improve Octomind to best serve your needs. Email me. Seriously.