Amsterdam gave us the usual November package: cold, grey, a bit of rain. But inside the JetBrains office it was warm, bright, and full of people who actually like talking about testing, which is all I really need. More than 40 folks showed up, grabbed some pizza, found their spots, and it immediately felt cozy in that very Amsterdam way.

The last year I saw a dog wandering around the office. This year the dog didn’t appear. Sad moment.

Still, the community made up for it. I saw familiar faces, new ones, and a few people who came up to me saying they want to give a talk next time. Two people even asked how to organise their own meetup. Honestly, this is the best feeling. This is how communities grow.

And I have to highlight this: last year Margarita came as an attendee. This year she came as a speaker. Helping people prepare for their first talks is one of my favourite things, so seeing that transition in real life is always a small celebration for me.

We had three talks this time, each with a different angle on testing.

AI QA: Non-Determinism in AI — A New Challenge for QA

Ivan opened the evening with a topic that keeps getting more relevant: how to test AI-heavy products when the system doesn’t behave the same way twice. His team works on coding agents, so they deal with multiple languages, IDEs, toolchains, workflows, and a huge number of execution paths. Basically a combinatorial explosion. He showed what happens when you try to apply traditional testing ideas to highly non-deterministic systems: assertions become useless, reproducibility disappears, and “expected results” sometimes don’t even make sense. So they had to rethink the whole approach.

The thing I kept thinking about during his talk: even before AI showed up, plenty of modern systems were already so complex that deterministic testing had quietly stopped working. AI just forced us to admit it. So it’s nice to see QA people experimenting with new methods instead of trying to make old ones fit.

A very grounded and honest talk.

Integration Tests for the IDE

Roman showed how JetBrains tests IntelliJ-based IDEs at the integration level. This included how they start the IDE, prepare projects, interact with UI elements, and hook into the internal APIs.

One thing I really appreciate is that JetBrains doesn’t just build features, they also build the infrastructure that lets them test those features properly. For example, they had to create a DOM-like abstraction over the IDE UI just to enable stable interaction. It’s a huge investment, but it makes a lot of things possible that most teams simply can’t do.

From Checklists to Jobs: Using JTBD to Guide Testing

Margarita’s talk was one of those talks where you see someone working through a real-world problem rather than repeating standard testing advice. She joined her team as the only QA, with different time zones, a lot of responsibility, and a huge amount of work. Instead of pushing harder, she stepped back and looked at her patterns through the Jobs To Be Done framework.

She talked about misunderstood requirements, unclear communication, fragile areas, long asynchronous workflows, and how all of this connects to the “jobs” she actually hires her testing tools and processes to do.

What I liked most is that she didn’t try to solve overload with "brute force" and overwork. She looked at her fears and frustrations not as personal weakness, but as signals that some parts of the process were fragile. That’s something many of us need to do more often.

Thank you, Amsterdam

To everyone who came — thank you. These meetups exist because of you.

Thank you to the speakers for preparing and sharing their work.

And thank you to JetBrains for hosting us again in their beautiful office.

If you want to give your first talk, or if you’re thinking about organising your first meetup, come talk to me. I’m always happy to help.

See you next time.