Dominik Kapusta

I found this really interesting, one of the more clearly articulated descriptions of how programming is changing, and the value of systems thinking https://lnkd.in/dsmwCbuP

11

1 komentarz

You’re the only QA on a brand-new project… what’s your first move? I’ve seen this situation many times - both in interviews and real life. A new project starts, and you’re the only representative of QA. It feels exciting, but also overwhelming. The mistake I see too often? 👉 Engineers jump straight into writing automation. 👉 Or they try to “cover everything” - even though there’s barely a product yet. It sounds productive, but in reality, it often wastes time. Here’s the approach that actually works: 1️⃣ Understand the domain and the customer. Spent the first weeks just learning how the business earns money. That gave us the context for deciding which features deserved the most testing effort. 2️⃣ Clarify requirements and expectations. Don’t assume the docs are complete - they rarely are. Sit with the PO, devs, and architects. Highlight the gaps early. 3️⃣ Build a lightweight test strategy. Think across all levels: unit tests (owned by devs), functional tests, non-functional (performance, security, usability). Even a one-pager test strategy creates alignment. 4️⃣ Evaluate resources and the team. Testing is never just “the QA’s job.” Developers own unit tests. Business owns acceptance criteria. QA connects the dots. 5️⃣ Define stack and approaches. Don’t pick tools just because “you know them.” Match the stack to the project. Sometimes Java + Playwright is perfect; other times Appium + Unity tooling is the only way. 6️⃣ Shift left. Test requirements and documentation first. Join grooming sessions. I’ve seen critical bugs avoided before a single line of code was written. 7️⃣ Lay down a skeleton for automation. Start small: CI integration, a few smoke tests, a clear folder structure. This becomes the foundation for future automation. ⚡ And the key reminder: automation is not a silver bullet. On some projects, the smartest decision is to keep everything manual for the first months. Rushing into automation without stable requirements usually creates more rework than value. 👉 The first months are not about “how many tests you can write.” They’re about building a foundation - so that automation, when it comes, actually supports the business. 💬 Curious to hear: when you join a project with zero QA, what’s YOUR very first step?

40

6 komentarzy

I had a great time presenting at Callstack's AI Meetup in Wroclaw, Poland, right before React Universe Conf kicks off. According to surveys, 60% of engineering leaders report no significant productivity gains from AI tools. Why? Developers only spend only 16% of their time actually writing code. Experienced developers will tell you, coding isn't the bottleneck, the issue is the other 84% of the job. In this talk I presented some stats about the current state of AI agents at work. I also introduced some practical tips for optimizing the non-coding parts of the job, like building presentations as React apps, automating technical documentation, and using MCP and APIs to connect AI agents directly to the tools we have to work with, like JIRA, Notion, or Github. The presentation itself was vibe-coded, which allowed me to build it quickly, demonstrating the advantages we can give ourselves as software developers. My slides: https://lnkd.in/exdVhyeY Blog post: https://lnkd.in/eUqHbBRT

35

2 komentarze

How do you know your app is ready to ship? When the core flow works end-to-end for a real user, without you explaining anything. If it solves one problem for one person, and does it well, it’s ready. The features, polish, and integrations can wait until after launch.

1

I remember spending an evening trying to write a bash script to verify a document signed through mObywatel — and not succeeding (mObywatel is a digital service provided by the Polish public administration). That was around the time ChatGPT started becoming popular. I wasn’t particularly motivated to solve it completely on my own, but with every new version of ChatGPT, I kept returning to the problem. The latest attempt, using ChatGPT-5 (I use ChatGPT Plus version), finally produced a working result. I’m sharing a brief write-up that focuses on the step-by-step verification scripts, which I believe are the most useful part. If anyone finds them helpful, the full post is here: https://lnkd.in/d4wmQx_A By the way, mObywatel remains somewhat underrated — it could serve as a solid foundation for interesting proof-of-concept projects around e-voting or digital signature collection.

11

3 komentarze

These are the skills no QA should ignore in 2025-2026. 82% of companies are already using or exploring AI. Yet, at our conference, TestCon Vilnius, only two out of 53 QAs had heard of LLM testing tools like Promptfoo or DeepEval. That’s the skill gap and the opportunity. In our January AI & LLM Testing Cohort, you’ll: ✅ Learn how to test models from Anthropic and OpenAI ✅ Validate them for security, hallucinations, and bias ✅ Build hands-on projects to prove your skills ✅ Get access to real hiring referrals for AI QA roles It’s the new QA skill stack for 2025. 🚀 Ready to join and become the next-gen AI QA Leader? Apply here → https://lnkd.in/e5nRzJ5k If you live outside the US, apply here for our special cohort for QAs outside the US: https://hubs.li/Q03Swlny0

29

1 komentarz