📰 Dev.to · Tudor Brad

Unit tests on a prototype. E2E on an MVP. Load tests on a feature nobody uses. Here's how we learned to match testing to the actual stage a product is in.

A migration that wiped 40,000 customer records, a checkout that double-charged on mobile Safari, and the uncomfortable pattern behind both.

We inherit broken test suites from new clients every month. Here is what went wrong, and the anti-patterns you can stop writing today.

Protractor is dead, Playwright works but the Angular-specific parts are fiddlier than anyone admits. Here's what we run, what broke, and what still makes us cur

We tried ML test generation, AI visual regression, and an LLM bug triage bot. Two of them wasted months. Here's what actually stuck.

What nobody tells you about being the lone tester on a scrum team, from the Friday afternoon asks to the PM sneaking in stories without acceptance criteria.

Automated tools catch roughly a third of WCAG issues. Here's what slips through, and the client audits that proved it the hard way.

A client got a legal demand letter over inaccessible forms. We ran the audit. 67 violations, most of them fixable in a week. Here's what we found and what it ch

A client wanted 80% automation in six months. We got to 45% and it was the right call. Here's the messy version of what an automation transformation actually lo

A junior QA hire automated the wrong things for six weeks. Here's what we learned about making people do manual testing before they touch Playwright.

We pointed a fuzzer at a client's API and found 14 crashes in the first hour. None of them were covered by existing tests.

Stripe test mode passed. Production failed on 3D Secure. Here's what we test and why.

Every time we inherit a client's test suite, we find the same problems. Hardcoded waits, brittle selectors, tests that test nothing.

Everyone agrees testing should start earlier. Nobody wants to invite QA to the design review. Here's how we handled that.

The test suite was green. The users were furious. Full automation sounds great until you try it.

Everyone says manual testing is dead. We have 50+ engineers and half of them do manual testing. Here's why.

We've inherited hundreds of test suites from new clients. The same mistakes show up every time.

A typo on the CEO's bio is low severity, high priority. A crash affecting 0.1% of users is high severity, low priority. Most teams get this backwards.

Your Cypress tests click buttons but never check what the API returns. Here's how to fix that.

We run both across dozens of client projects. Here's when we pick each one and where both fall short.

Nine spreadsheets, two conflicting reports for the same step, and screenshots buried in Slack. That's how Auditi started.

Regression tests in staging passed. Production broke anyway. Here's why 'test it in staging' isn't a QA strategy.

Performance, security, accessibility, usability. Everyone agrees they matter. Almost nobody runs them until something breaks.

AI is writing code faster than ever. The question nobody's answering well: who makes sure it actually works?