Expert testing for AI agents and apps. Helping businesses ship with confidence since 2015.
Ukraine
Joined April 2016
- Tweets1.6K
- Following2.3K
- Followers491
- Likes695
1/9 Every few months a new article drops declaring AI has finally "solved" software testing. 2024, 2025, now 2026.
At QAwerk, we've spent the last year actually running these tools on real client projects. Not benchmarks. Real codebases, real deadlines, real bugs.
8/9 Ten years running QAwerk, we've watched every wave of "this will automate QA." Codeless testing, record-and-playback, model-based testing. Each one moved the floor up. None of them replaced judgment.
found a bug last week that passed every test
AI wrote it. looked clean. compiled fine.
failed in production because it had no idea how the rest of the system behaved
green tests aren't the same as correct code.
1/9 Every few months, a new wave of "how to become a QA automation tester" guides floods the internet. Coursera, Udemy, YouTube. take your pick. The roadmaps are always the same: learn Python, learn Selenium, learn some CI/CD basics, profit.
8/9 - Being comfortable with ambiguity. especially when requirements change mid-sprint
- Communication skills, because a bug report no one reads is just noise
1/8 There's a pattern we keep seeing on teams that adopted AI coding assistants this year.
Developers are shipping features faster. Sometimes dramatically faster. And the test coverage... stayed exactly where it was six months ago.
7/8 None of this means AI-assisted testing is useless. We use it. It speeds up the scaffolding, handles repetitive regression scenarios, and is genuinely good at certain kinds of coverage. But someone who understands what "done" actually looks like has to be in the loop.