On the craft of software testing — accessibility, AI & testing, ethics, technique, and career.
Filtering as you type — press Enter for a full-text search.
AI can now diagnose a flaky test and open a pull request to fix it, but a repair that changes what the test checks and one that just silences it produce the identical green tick — so every automated fix is a candidate you review, never a result you trust.
The claim that manual QA is dead is clickbait — what's dying is manual test execution, while judgment, exploration, and test strategy become more valuable, so the move is up the stack, not a career restart.
The threat to QA in 2026 isn't a robot taking your job — it's three pressures at once (AI mandates, tighter budgets, and a flood of AI-generated code) that make human judgment more valuable, not less.
AI writes more tests with more assertions, but assertion count isn't the same as bug-catching — the only test worth keeping is one that can fail, so reviewing AI-generated tests means confirming they'd actually notice a broken behaviour.
Nearly nine in ten QA teams are experimenting with generative AI but only about one in seven have scaled it, because the hard part was never the model — it's the ownership, data access, and trust required to run it in production.
A flaky test isn't random — it's a deterministic bug whose trigger doesn't fire on every run, and the trigger is usually a timing or waiting problem rather than bad luck or a bad selector.
Code coverage measures which lines your tests execute, not whether those tests would catch a bug. Chasing a coverage percentage produces tests that run everything and verify nothing — here's what to measure instead.
If your team treats testing as a formality, the fix isn't testing harder — it's making the value of the bugs you catch legible in the language the business already cares about: risk, cost, and time.
Shift-left testing means finding defects earlier by involving testers in requirements and design, not moving all the testing work onto one person. Here's the difference — and how to tell which one your team is doing.
The AI breaking your software and the AI testing it are the same story. Why verification — and a real tester in the loop — is where the value now lives.
Where AI genuinely helps in testing, and the guardrail for each — because the same tool that speeds you up will hand you confident, wrong, green results.
What tester forum threads reveal that analyst reports never will — and how to read them for observations, not conclusions, when deciding your QA career.
No — AI changes the tester's job, it doesn't remove it. What AI can genuinely do in testing, what still needs a human, and where the line falls.
A bug report is a persuasion problem, not a form. The checklist that gets bugs fixed — and the one thing that matters more than all of it: reproducibility.
Automated accessibility tools catch the mechanical failures and miss the judgment calls. What each one really finds, and why the answer is "both, in the right order."
A test strategy is the how across the org; a test plan is the what for one project. The real difference, in a table — and when a small team actually needs each.
A radiation-therapy system's documented limitation still killed eight people. On the distance between "documented" and "safe", and testing for the confident wrong answer.
Accessibility testing starts with your keyboard, not a scanner. The first real checks any tester can run against WCAG 2.2 — and where the tools quietly stop helping.
How a one-second clock correction took down Reddit, Cloudflare and more — twice — and what leap-second bugs teach about the assumptions software never tests.
Pseudolocalization — a fifteen-minute stress test that breaks your international UI before a single string is translated, catching hardcoded strings, overflow and encoding bugs early.
No articles match that filter yet.