Welcome to the 320th issue!

Everyone is trying to improve testing with AI agents.

But how does it actually work at scale?

Here's one answer: Agentic Testing at Slack: Where Agents Fit in the E2E Testing Stack.

Big kudos to Sergii Gorbachov for describing it in such detail.

Happy testing!

Dawid Dylowicz

Daria Kotelenets shares a practical framework for startups that focuses on three things: making fast and informed decisions, removing low-value work and building a whole-team approach to quality.

However, Keith Klain warns against Testing at the Speed of Stupid in the age of AI.

If you work on LLM evaluation, Giorgos Siamantas did an interesting experiment comparing two different models, showing that even a model scoring 1.00 can still miss what the reader actually needs to know.

Someone on Reddit opens up about the burnout of constant task switching after their company pushed shift-left testing. If you feel similar, there's some good advice from the community on how to handle it.

Also, Jesper Ottosen contemplates: What is next after an experienced quality role?

In case you missed it, there's a new HTTP QUERY method! It works like GET but supports a request body, which fixes a long-standing problem for search and filter endpoints.

You can also learn more from Manish Dhakeria on Why QA Engineers Should Care About the New HTTP QUERY Method and from NaveenKumar Namachivayam showing how to use QUERY method in Node.js, curl, k6, and JMeter.

What skills matter when building a quality team? Gary Hawkes shares how hiring for personality alongside hard skills helped grow a diverse, self-navigating team from two to thirteen people.

Furthermore, Dayana Lanova shares insights into What 5 Years of QA Taught Me About Building Better Products, while Dilshi Piyumika outlines What Truly Makes a Great Software Tester?

How do you decide which flaky tests to fix first? Drexel McMillan proposes a risk-based formula for scoring test automation failures and deciding on the priority. You can read more in the second and third parts.

A/B testing and feature flags are useful, but they can also disrupt testing, as Viktar Karanevich found out. The fix? Focusing on testability first so that tests can force deterministic states.

Wondering how to build a scalable test automation solution from scratch? Laxminarayana Boga put together an insightful series of 11 articles explaining the approach step by step using Playwright, BDD and more.

Also, Sébastien Gineste advises to Stop Mocking Everything: The 3 Tests I Start With for a Service.

Vitaliy Potapov shares how he uses BDD with coding agents to plan features instead of markdown specs. The scenario diff works as both a review artefact and a test.

On that note, Daniel Sayer explains why Your AI agent doesn't care how you write acceptance criteria. Here's the experiment that proves it.

An interesting approach by Dmytro Pishchukhin, who built a Claude Code plugin with three skills that run your integration suite autonomously or guide you through manual steps and write a protocol file every time.

Also, Abhijeet Vaikar points out that Your agents are happily automating an untestable system.

If you run Cypress inside an AI agent, Emily Wisniewski has a quick tip to cut the token cost of test output by three-quarters with two flags.

Moreover, it's worth knowing that The Electron browser is being deprecated in Cypress.

Eliminate flaky Playwright tests with a resilient, factory-based Page Object Model. In this step-by-step guide, Marcin Michalak demonstrates how combining TypeScript Getters with an asynchronous init() pattern enforces clean test architecture.

You can also learn from Dmytro Budym on how Playwright 1.61 quietly gave the trace viewer visibility into WebSocket traffic.

If you wonder how to use AI agents to help you test an app, Debbie O'Brien built a solution that finds bugs by exploring the UI and creates bug reports. Another agent then reproduces and fixes these bugs, creating a pull request.

Got fed up with Appium? Amr Salem shares Astur — a new framework for native mobile platforms that gives you a familiar Playwright API for Android and iOS.

Have you heard of Webwright? It's not a test framework but a code-writing agent that produces reusable Playwright scripts from a plain English goal and a URL. Dinesh Gujarathi shows how.

Want to learn how to test AI-powered apps? Jason Arbon, who has worked with AI for a few years, wrote a practical book covering evals, sampling, RAG, agents and prompt injection.

The way of working with AI is changing quickly. Joe Colantonio invited Matt Wynne for a thought-provoking talk about how the tester's role can evolve towards system validators with the help of AI agents.

In this 21-minute recording, Daniel Knott gives good advice on setting up test automation practice in companies, outlining the foundations and challenges.

Trying to prove it works... 😂

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