TestMu AI Blogs

Stephan Mousset of ING on golden signals, choosing tests by risk instead of habit, workload modelling from production, and closing the feedback loop.

Shafik Quoraishee of The New York Times on running agents on the device, the space, frame and energy budgets they fit inside, and active accessibility testing.

Six QE leaders were asked what they killed to fund their AI bets. None named anything, and their reasoning is the most useful part of the panel.

Shantanu Wali of TestMu AI on why testing agents drive browsers not phones, Kane CLI on simulators, and assurance that runs from requirement to proven coverage.

Thejes Sree Satheesh Kumar on turning an OWASP LLM risk into a check that fails the build, why step four stalls teams, and sampling instead of pass or fail.

Rahul Parwal on moving from execution to insights, publishing what you learn the same day, and why who knows you matters more than who you know.

Rashi Agrawal of Hinge Health on going all-in on agents, the QA backlog that vanished, and why the bottleneck moved upstream to judgment about what to build.

Dona Sarkar on why AI is stuck in its scroll bar moment, the three kinds of software nobody knows how to test, and the faster horse test.

Manideep Singireddy on scenario-first test data: derive the scenario from test intent, then create only the data it needs. Volume is not coverage.

Sachin Sharma on why end-to-end testing still feels manual: every box of the testing lifecycle is automated, and the arrows between them are not.

Pair an iPhone over USB once, then automate it entirely over Wi-Fi from macOS, Linux or Windows with no Xcode. Inside the appium-ios-remotexpc workshop.

Replit's CTO on the agentic loops behind a claimed tripling of output, why verification sits inside every one, and what replaces tests outside engineering.

Laveena Ramchandani on why a green check mark is not a reviewed change, what stays human, and naming an owner for every automated decision.

Vibhor Rastogi of Citi Ventures on why one license can run a thousand agents, what that does to unit economics, and why outcome pricing still has no referee.

Pramod Yadav of ABN AMRO Clearing walks a full continuous delivery pipeline, agentic semantic versioning inside the PR, and one commit behaving three ways.

Five enterprise quality leaders on why pilots stall: capacity is no longer headcount, autonomy must be earned through evidence, and testing becomes trust work.

Walter Zimerman of Amazon on why AI fails confidently, how accuracy compounds downward across chained agents, and what evaluation-driven design fixes.

Four SDET skills you already have, four to stack on top, and the artifact they combine into: an eval, a test suite for a system that never repeats itself.

Deepshikha and Anamika Mukhopadhyay of Nagarro on five ways agents fail in production, and how to test planning, memory, tools, identity and communication.

Mathias Strziga on why a red test result is a classification problem rather than a bug, and where the human gate sits in an agentic QA pipeline.