Welcome to the 333rd issue!

This discussion caught my eye:

My company is killing off the junior QA role, and I fear for the whole QA department.

A senior QA at a fintech startup shares their observations. First, the manual tester role went. Now junior roles are going too, as developers and AI take over the "easy" work.

On the other side, a senior developer wants to move into QA to audit AI-generated code, and asks: Is QA actually dead or is the market missing something?

Like you, I'm curious where this goes next.

For now, happy testing! 🙂

Dawid Dylowicz

Has testing become the bottleneck since your developers started using AI? Martin Cáceres González says we should use it for routine work and invest our time in deciding where a change can break.

Moreover, Vineet Verma makes a similar point: AI made code faster to write. Nobody made it faster to trust. That gap is landing on QA's desk.

Madhawa Ratnayake shares 10+ lessons from testing with coding agents, such as how to check the test data before you trust a failed test.

At the same time, Emre Gültekir reminds us that A Green Test Only Answers the Question You Asked.

Thinking software was better before AI? Here's a fun read from Jacob Voytko on how often software crashed in the 90s. So even with AI writing more code, the software we test today is in better shape.

Filip Hric makes a case for testers to become technical, starting with how AI agents work and how to evaluate them.

On that note, the Reddit community discusses: Will SDEs gradually take over SDET roles as AI tools become more capable?

Many outages start with a change to a live system. Ben Congdon describes how AI agents can now check each release and detect problems better than standard alerts.

Similarly, Sourik Ganguly shares How to Verify AI-Generated Code: Checks, Review and Receipts in Accension.

Many of us live with one huge test suite because splitting it used to cost too much time. Kavitha Namduri says AI makes that work cheaper and suggests where to start.

Furthermore, Mohammad Faisal Khatri gives a good overview of The Test Pyramid in the Age of AI-Generated Code.

Do your tests often pass after a retry? Nathan Broadbent shows how to keep a record of those failures and give it to a coding agent to fix.

Moreover, Esra Kocaoğlu shares What Happened When I Let an AI Agent Fix My Failing Tests.

Ahmet Kalafat demonstrates how he built two-stage validation into a GenAI pipeline where deterministic rules make the decision and the LLM only advises. You can also read parts 2 and 3.

Speaking of that, Swati Seela describes 11 Mistakes We Made in Test Automation That We Shouldn't Repeat With AI Judges.

Coding agents can write lots of tests. But which of them would catch a bug? Valentyna Polienova outlines how they audited a Flutter test suite and turned the lessons into two skills for agents.

Also, Daniel Seidewitz writes about a case where My Coding Agent Passed Every Test. The Code Was Wrong.

A helpful series of articles by Moshiur Rahman on how to get useful Playwright test drafts from an LLM. You can already read the following parts 2, 3, 4, 5 and 6.

In relation to that, Vineet Verma explains going From 70% to 99% Pass Rate: Why Playwright's New Isolated Retries Make Flaky Tests Easier to Read.

Using AI to write your end-to-end tests? Indra Aristya gives good advice on what to check next so that you can trust your test suite.

What's more, here's a positive story from Tayguara Dias Reis: I let an AI agent write my Playwright API tests. It caught a bug my suite missed.

That's a helpful comparison from Myke Ralduxin of the popular AI coding assistants for testing, including a simple plan to try them on your own test suite.

On top of that, Ibrahim Denis Fofanah points out: Your Agent Tests Pass. They Mostly Test the Plumbing.

Wondering about Vibium? Boni García shows how to let a coding agent test its own work in a live browser and use a second AI model to verify the feature.

Similarly, Volta Jebaprashanth shares a related approach: I Stopped Reproducing Failed Selenium Tests. Now the AI Debugs Them in the Browser They Failed In.

What happens when developers are responsible for testing their own features? They get creative. Anurag Kumar Singh describes how they built a small device lab that runs 300+ Maestro flows every night.

Also Ravi Rawal tells us How to Apply SOLID Principles with Maestro and How to Organize Maestro Tests.

Git is easy... 😂

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