Welcome to the 318th issue!

This week, I want to highlight a talk by Niklas Gustavsson, Chief Architect and VP of Engineering at Spotify:

Scaling Developer Experience to Teams and Agents at Spotify.

It's a wonderful sneak peek into how big companies build custom agentic AI solutions to accelerate software development.

The challenge? Like probably most of us have already experienced — nearly double the volume of PRs to review.

Enjoy and happy testing!

Dawid Dylowicz

Good read from Ishalli Garg on how testers are often the first to spot when a product stops serving users and starts serving its own architecture, including advice on what to do with it.

Moreover, Gojko Adzic interestingly states that In five years, everyone will be a product manager.

Will AI replace testers? Alona Titova doesn't think so, having lived through Selenium and record-and-play automation. It seems the pattern repeats now with AI, too.

Also, worth learning from the community: Did AI actually improve your efficiency?

Dan Ashby makes a strong case that quality is an emergent property of how well your whole engineering system is coordinated, rather than something done at the end of the pipeline.

Also, Rajeshkumar Rajaseakaran Nair points out Why Good Test Engineers Think Like System Architects.

Do you feel like testers are sometimes seen as less valuable than developers? Vincent Ferreira breaks down why that view is outdated and shows how to make your QA work visible and respected.

On that note, Ferawati Hartanti Pratiwi writes about Reframing Quality: Everyone Owns Success, QA Owns Failure — Everyone Owns Quality.

Anastassia Voronina makes a case for treating the quality gates, not the code, as your source of trust by providing an AI dev team with linters, test baselines and fresh-eyes reviews.

What's more, Chris Kenst shares some thoughts on AI and the Declining Cost of Trying.

Continuing the series started last year, Lena Pejgan Nyström shares a useful perspective on what it takes to step back and see the bigger picture when the team has stopped questioning why things are broken. You can also read parts 4, 5 and 6.

I also found this discussion interesting: What are the most frustrating and time-consuming parts of your QA job?

Want to use AI for automation? This is a good warning on how it keeps introducing things like testing happy paths only, mocking entire modules or ignoring existing fixtures.

Moreover, Nael Marwan advises to Stop Guessing Whether Your Test Is Flaky. Run It Five Times.

How would you test a mobile app that has 30+ languages and thousands of UI permutations? Asa Kusuma and Chad Hietala from LinkedIn explain how they use a QA agent with vision-language models to find bugs without relying on selectors.

Also, Sourojit Das put together a handy guide to Building Your First AI Agent as an SDET, while Medhavee Upadhyaya shares a story of how I Ran My AI Testing Tool on Its Own Code. It Found Bugs I Missed.

Is automation testing a dead end? Lilia Urmazova thinks the role is changing due to AI and believes that testers need to move up to architecture, system design and business analysis to stay relevant.

For that, Rajesh Yemul started a comprehensive The Complete Agentic AI for Quality Engineering Series with more than 15 parts and going.

Tuấn Huỳnh makes a good case for wrapping raw locators in typed classes that give type safety, self-documenting page objects and one place to update when something changes.

This is followed by Tuấn's article on Self-Healing Locators with RAG and Vector Databases.

When it comes to migrations, unit and integration tests alone may not give you the full confidence. Enrico Piovesan explains why and suggests doing parity testing to catch the bugs that other tests were never designed to find.

If you're automating mobile apps, you may like this. It's an AI agent that lets you test iOS and Android apps with tap, swipe, profile and debug on simulators directly from the CLI.

Wondering what changed in Playwright and Vibium over the last 6 months? Beth Marshall went back to an old code repo and tracked 20+ releases of each tool to demonstrate how they moved toward agent-driven browser automation.

And speaking of tool comparisons and updates, Vince Graics shares that WebDriverIO Went Tracing.

Running BDD at scale requires good planning. Laxminarayana Boga walks through how to structure fixtures, page objects and tags to keep a 3,000-test test suite manageable, followed by Highly Scalable Test Data Management in Playwright.

Viranga Bandara shares how adding a functions layer above standard Page Objects was the key change that let a big test suite grow efficiently.

He also followed up with an article on 5 Selenium Anti-Patterns Found in Production Frameworks — And How to Fix Them.

A helpful article on what Playwright screenshot in CI means and why adding waitForTimeout hides the problem rather than fixing it.

Moreover, Razvan Vancea advises trying to Fail Fast in Playwright with maxFailures.

An interesting, provocative 16-min talk by Keith Klain on how the longstanding promises of improving quality and reducing cost with tools and automation didn't quite play out as planned.

HTTP codes... 😂

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