TestMu AI Blogs

Two UKG engineers on scoring every test against each code change, the six signal categories behind the score, and why they still recommend periodic full runs.

Four QE leaders on how much of their code AI writes, why reviews now outlast writing, and the failing test that turns out to be the test's own fault.

Hammad Ahmed on the seven-agent QA swarm behind a decision intelligence platform, why bigger context made it worse, and how skill gating fixed it.

Jyotheeswara Reddy Gottam on quality memory, why generated tests must earn their place, and measuring outcomes rather than how busy an agent looks.

Prashant Sharma of Arm on why build once, run everywhere is the wrong model for AI, and what a developer-defined quality envelope has to prove on each target.

Haritha Sreedharan Nair on building your own agent benchmark: grade the trajectory rather than the pass rate, and write evals from the user's job.

Srinivasan Sekar and Sai Krishna on the official Appium MCP server: 33 intent-shaped tools, ranked locator generation, and why the agent is not the MCP.

A Testμ 2026 retail panel on continuous readiness, bot traffic already being blocked, and why agents test the data layer rather than the storefront.

How Fox gates QA agents behind an eval suite written first, the translation attack that leaked a full system prompt, and why eval coverage is permission.

Anubhav Singhmaar on why a green check mark is a claim rather than proof, and the sealed .evidence pack built to carry the record of a test run.

Application, model, infrastructure and data - Maryia Tuleika on the four layers of AI security, and why testing only the one you can see gives false confidence.

Shravan Mahajan runs one PRD through Kane CLI to a passing iOS simulator test - four use cases, seven acceptance criteria, two tests, one executed.

Siddhant Sinha runs a PRD through Kane CLI end to end: use cases, acceptance criteria, wired tests, a real browser run, and coverage backed by evidence.

Vanessa Johnson of The New York Times on annotating a GraphQL schema with accessibility directives and generating Playwright and axe-core tests from it.

Four delivery leaders from Wipro, Capgemini, Nagarro and Persistent on guardian agents, the new QE pyramid, and where humans still sit in the loop.

Three healthcare QE leaders on signing off the protocol instead of the artifact, why synthetic data deletes edge cases, and controls that are only sentences.

Lazar Jovanovic of Lovable on going way past vibes: skills instead of prompts, design systems, project knowledge, managed secrets, MCP both ways, and shipping.

Justin Roy of Microsoft on the five questions that decide whether people can work with an agent: can I see it, understand it, influence it, own it, trust it.

Safeena Banu on architecting agentic QA - six layers built bottom-up, tools as API contracts, and why the goal is the least autonomous system that works.

Athresh Guruprakash on rescoping Selenium rather than retiring it, the three layers of test self-healing, and why the human approval gate stays.