Welcome to the 324th issue!

This discussion caught my eye this week:

Is dedicated QA still a real career or am I chasing something that's disappearing?

Mind you, QA was declared dead many times in the past, though somehow it's still here.

But it's natural to worry and this thread gives a glimpse into what the testing community really thinks.

Also related — What is happening in your companies within the AI era?

Happy testing!

Dawid Dylowicz

Some repetitive testing tasks can be daunting. Iraida Chirkova shares which of these parts she managed to delegate to AI, while keeping the exciting stuff for herself.

Moreover, Kristin Jackvony has good advice on drawing that line in How to Use AI With Your Brain.

Any existing automated ways of testing are the primary target for AI. But testing is more than that, and Martin Ivison shows how our job splits into three parts and where we're still safer than AI.

Moreover, Dmytro Shyshkin asks: Will AI Replace QA Automation Engineers? — to which Shoaib Ahmed Quraishi has an answer: Every QA Engineer Will Need to Think Like an AI Engineer Within 5 Years.

Jason Arbon started a great series of articles on how QA leaders should respond to AI doing a significant part of testing on its own. This continues in parts 2, 3, 4 and 5.

Also, Melissa Fisher takes this to the process side in Quality Management System has the same parts but how it is operating is changing.

Mårten Schultzberg from Spotify breaks down their lessons learned on LLM-based A/B tests. Turns out that while it can work, it requires strict conditions and calibration.

At the same time, Keith Klain wrote a good piece on The Verification Asymmetry Problem.

Julia Pottinger shares how testers can work with coding agents and breaks down which tasks are safe to hand over. This is followed up by an article on Working With AI Coding Agents as a QA Engineer.

What's more, Johnny Janzen shares the process of building Agentic Continuous Delivery.

Can you make AI test mobile apps? Rashiprabha Atapattu asked that question and built a skill that checks for a live emulator, reads each pass/fail mark, and takes a screenshot on failure, followed by the final part on Where AI Still Needs Human Judgment.

Birgitta Böckeler put TDD to the test with AI coding agents and, interestingly, found no strong sign that it improves design, test quality or mutation scores over skipping it.

Moreover, Sławomir Radzymiński takes a closer look at that measure in Mutation Testing for Agent-Written Code.

How useful is test coverage as a metric, really? Jitesh Gosai says that real confidence comes from monitoring, rollout control and how fast you can undo a change.

On that note, Josphine Job suggests a different check in Invariant Testing Solves the Oracle Problem in AI Testing.

AI can produce code that passes tests yet still solves the wrong problem. Dennis Martinez shares three things to look for in AI-written tests before you trust the code.

Additionally, Anton Gulin has a longer checklist for that in How to Review AI-Generated Tests: Seven Checks Before You Keep Them.

Automated visual tests often fail without explaining why. Martin Poole shares a practical approach using GitHub Actions, Playwright and a vision model to add useful context to every report.

Also, Gil Zilberfeld asks what AI does to repeatability in AI vs CI: Nobody Told the Pipeline.

Interesting announcement from the CEO of Postman, preparing for a world run by AI agents. Not affiliated in any way, just highlighting the direction that some vendors are taking, using AI as leverage to expand beyond their proven capabilities.

Cypress Cloud suffered a security incident through third-party tool Metabase, exposing some account data, tokens and test data. Here's the official response and recommended actions that you should check now.

Also, Emily Wisniewski describes a new feature: Cypress tap: give your AI agent the context the Cypress app shows you.

Svetlana Tretjakova lists simple rules to make Playwright tests less flaky, from waiting for actual conditions to avoiding shared test data and using traces to debug failures.

Svetlana also shares a debugging trick in How I Use Custom Playwright Fixtures to Log Failed API Requests Automatically.

If you want to mock, block or slow down API calls in your Playwright tests, Mohammad Faisal Khatri shares a practical guide to network interception, covering request mocking, HAR replay and the Trace Viewer.

Speaking of that, Viranga Bandara covers the Cypress side in Mastering cy.intercept(): Mocking APIs and Testing Edge Cases with Cypress.

Wondering how to test AI? This hour-long discussion between Joe Colantonio and Jason Arbon gives plenty of insights into how to approach it.

Especially important since AI Is Shipping Code Faster Than QA Can Keep Up, as Alex Khvastovich explains.

And testers face both... 😂

Thanks for reading!

If you enjoy this newsletter and find it helpful in becoming a better tester, please consider sharing it with others.

Dawid Dylowicz