Writing
Which technique finds which bug, how to catch a requirement that cannot be tested, and where AI earns its place in a QA workflow.
Writing the test first proves it can actually fail before you trust it to catch a regression. Here is the red-green-refactor cycle, a worked example, and the cases where TDD stops paying for itself.
Chrome DevTools MCP hands an agent a live browser: console, network, performance traces. This post separates what that unlocks for QA automation from the parts of a real bug report it still can't produce on its own.
Playwright MCP gives an agent the same browser control a human tester has, without the agent writing raw Playwright code by hand. Where that trade holds up, and when Chrome DevTools MCP is the better fit instead.
Most teams reach for an MCP server backward: they let the agent click around first and write assertions after. This post lays out the order that holds up in review, case first and agent second, and shows where Playwright MCP and Chrome DevTools MCP diverge on what they can assert.
A one-line fix can still touch a dozen call sites you never opened. Impact analysis traces a change to what it edits, depends on, and shares data with, so the regression suite you rerun matches the actual risk instead of the whole product.
An AI test case generator excels at the input you hand it and struggles with the edge cases you didn't think to mention. This post adds two sections on where that gap shows up and how the output holds up once it lands in test case management and a real QA automation pipeline.
Equivalence partitioning breaks down once two or three rules interact at the same time. This post works through a decision table and a state diagram side by side, so you know which one to reach for when a bug turns out to be an illegal move instead of a bad value.
Two model releases landed days apart in September, and each one changes a different part of a product team's workflow. What each model handles better, and three habits worth adjusting this release cycle instead of waiting for the next one.