Hello there! đ
Welcome to the 229th edition of Software Testing Notes, a weekly newsletter featuring must-read content on Software Testing. I hope this week has been good for you so far.
This week:
Why Testing AI Agents Is More Conversation Than Code
The QA Blind Spot Nobody Talks About
Give Your AI Agent Eyes: Capturing DOM Snapshots for LLM-Assisted Test Debugging
Testing Google ADK TypeScript Agents
Beyond Load Testing: Rethinking Locust for Real User Workflows
and moreâŚ
⨠Featured
No vendor lock-in: Build tests visually. Own them as YAML.
As a QA professional, you shouldnât have to choose between fast test creation and control over your test assets.
With BugBug, you can now record tests visually and export them as human-readable YAML. Keep them in Git, review changes, create backups, edit them outside the platform, and import them back anytime.
Your tests stay portable, reviewable, and under your teamâs control. With no vendor lock-in. See how BugBugâs test import & export works.
đ Testing
QA in a Vibe-Coded World: What the Discipline Needs Right Now by Joshua Bihun
AI is taking more of the mechanical work out of QA, but that makes good judgement more important, not less. Joshua Bihun looks at where AI can already help with things like bug reports and test data, and why testers also need to get comfortable testing AI itself.
Breaking what you thought you knew about testing in 30 minutes by Maaret Pyhäjärvi
A CV can tell you how long someone has been testing, but exploring an unfamiliar app can reveal how someone actually thinks. Maaret Pyhäjärvi uses that exercise to challenge test-case habits and see how candidates learn, investigate, make sense of a domain, and respond to feedback.
Why Testing AI Agents Is More Conversation Than Code by Gayathri Bolineni
You can't test a conversational AI like a traditional application when the same input can produce different answers. Gayathri Bolineni looks at testing the outcome instead, including multi-turn context and failures across NLU, RAG, APIs, and guardrails.
Quality falls through the gaps by Jitesh Gosai
A team can have plenty of confidence in its own component and still ship a broken user journey. Jitesh Gosai looks at what gets lost between components and teams, and how integration, journey, and system levels can expose those gaps.
The QA Blind Spot Nobody Talks About: Testing for Emergent Behavior by Milan Bomjan Tamang
Testing individual AI agents wonât necessarily reveal what happens when dozens of them start interacting. Milan Bomjan Tamang looks at testing agent swarms at scale, including shared communication channels, collective behaviour, and independent verification of activity.
Building a quality culture from zero by Callum Akehurst-Ryan
Building a quality function isnât simply a matter of introducing better testing practices. Callum Akehurst-Ryan describes how coaching, hands-on testing, shared signals, and reusable tooling can gradually move quality into the teams themselves.
Are Test Management Tools Becoming Obsolete? by Emna Ayadi
Emna Ayadi looks at whether test management tools are becoming more trouble than theyâre worth. The answer may come down to whether they genuinely improve traceability and confidence.
You Cannot Test Everything. So What Deserves More Attention? by Aaron Boadi
When testing time is limited, running more tests isn't necessarily the answer. Aaron Boadi argues for continually asking what the team still doesn't know, then putting attention where uncertainty and consequences are highest. That makes prioritisation less about clearing a regression list and more about learning what could actually hurt the product.
đ Software Testing
âď¸ Automation
Give Your AI Agent Eyes: Capturing DOM Snapshots for LLM-Assisted Test Debugging by Irfan MujagiÄ
When an AI-generated Playwright test fails, a timeout message doesnât give the agent much to work with. Irfan MujagiÄ shows how capturing a screenshot and ARIA snapshot alongside the failure gives an LLM useful page context.
Testing Google ADK TypeScript Agents Without Chasing Sentences by Raju Dandigam
If an AI agent gives the same answer every time, you're probably testing the wrong thing. Raju Dandigam shows how Google ADK tests can focus on decisions, tool calls, state, and safety boundaries.
Implicit Assertions Are More Readable by Gleb Bahmutov
Cypress API tests don't need a separate explicit assertion for every property in a response. Gleb Bahmutov shows how its().should() and cy-spok can express checks for nested values, types, patterns, and ranges more compactly. The result is less boilerplate and tests that make the expected response structure easier to see.
đ¨ Performance
AI Performance Testing Guardrails Every Tester Needs by David Mello
AI can write a performance test that looks convincing while quietly hiding the failures that matter. David Mello lays out guardrails around specifications, deterministic scripts, rules files, and independent review to keep that from happening.
Beyond Load Testing: Rethinking Locust for Real User Workflows by Priyanka Kannan
A system can handle thousands of users and still struggle when the real problem is the amount of data being processed. Priyanka Kannan uses Locust to model complete user journeys across different dataset sizes and compare performance against baselines.
đ Accessibility
Lighthouse 100 Does Not Prove AccessibilityâHereâs What to Test Next by Bertrand Morel
Accessibility Testing can be tricky and completely relying on automated tools is not a good idea. Bertrand Morel looks at what automated audits miss, from keyboard interactions and focus management to assistive technology and complete user journeys.
đ ď¸ Resources & Tools
Playwright + AI Testing Lab â Hands-on Playwright automation lab showing framework evolution from vanilla tests to fixtures, authentication, API testing, and isolated AI-assisted testing with Midscene.
đ Software Testing Creators
Do you create content about software testing? Add your blog, website, or newsletter here, and Iâll add it to the list.
đ¨ Have Something to Share?
If youâve written an article, tutorial, built a useful tool, or found something the testing community should know about, send it my way via DM on X or submit it here.
And if you spot a mistake or have an idea for improving the newsletter, just hit reply.
đ Other Interesting Stuff
â Last Weekâs Most Read
Stop Reporting Test Activity: Show the Business Impact of Testing Instead by Richard Forjoe
Apparently We Need a Testing Mindset After All by Keith Klain
What Is the Right Question a QA Should Ask During Refinement â Especially in the AI Era by Marina JordĂŁo
đ And Finally
The stages of the QA Cycle đ¤Ł
Thanks to everyone for subscribing and reading!
Happy Testing!
Pritesh
@priteshusdadiya