Notes from building software that gets inspected.
Agent architecture, evaluation, and the test engineering discipline underneath it all. Written by the people doing the work.
Playwright MCP does not heal your tests. It removes the thing that breaks.
Playwright MCP drives the browser through the accessibility tree, not selectors. That removes what self-healing was patching, and moves the problem.
The hardest part of automating regulatory reporting is the list
Before choosing a system, write down every return you owe. What a return registry holds, what automation adds, and what breaks when a template changes.
There is no seven-day STR deadline. It was amended out of the Act.
Section 7(1) of the AML Act once gave you seven working days to file an STR after forming suspicion. That wording was substituted with "promptly". CTRs kept the seven days, which is where the…
Phase I of STR-F automation is done. Phase II is the one that will hurt.
FMU Circular 04 of 2025 moved intra-bank counterparty transactions to goAML XML. Phase II extends that to inter-bank, and the banks that wait for a date will be the ones scrambling.
The CBUAE wants an inventory of every AI system you run. Here’s what that takes.
The UAE Central Bank's February 2026 AI guidance asks for a model inventory, annual bias testing and accountability for third-party models. Three obligations that take months, not weeks.
Your AI Testing Pilot Is Going to Fail. Here’s Why
Most AI testing pilots stall out — and usually not because of the tools. A look at why these pilots fail, and what separates the ones that actually stick.
The Post-Code Era: Is Project Management Obsolete?
When AI writes the code, the bottleneck shifts from building to deciding what to build. What that change means for how we plan and manage software projects.
We Built an AI Agent That Runs Social Media Autonomously
Not another template filler — a fully autonomous AI agent with a ReAct reasoning loop, 15 tools, rejection memory, and auto-publishing. Here's how we built it and what we learned.
Why QA Teams Are Secretly the Best AI-Assisted Developers
Everyone's talking about AI replacing developers. But the teams getting the most out of AI coding tools aren't dev teams — they're QA engineers who already think in edge cases, validations, and test-first workflows.
Playwright vs Cypress in 2026: Which Should You Pick?
We've shipped production suites in both. Here's an honest comparison based on real projects — not docs and marketing pages.
When Should a Startup Invest in QA? (Hint: Before You Think)
Most startups wait until something breaks in production. By then you've already lost users and trust. Here's the right time to bring in QA.
Embracing AI in Software Testing: Navigating the Evolution
AI is reshaping how we approach testing — from test generation to intelligent defect prediction. Where the industry is headed and what QA teams should do now.
Rethinking Self-Healing Test Automation
How self-healing locators work, the three costs nobody logs, and what the idea should mean instead: locate by role and name, and let the build go red.
Testing Fintech Apps: What Actually Breaks in Production
After testing payment apps, lending platforms, and earned wage access systems — the patterns that break repeatedly and how to catch them first.
The CATS Principle: Building Financial Software That Deserves Trust
Compliance, Accuracy, Traceability, Security — four pillars that every fintech QA strategy should be built on.
Cypress Test Debugging: Terminal Report Deep Dive
A hands-on guide to debugging Cypress test failures using terminal reports — the patterns, the gotchas, and the workflow that saves hours.
WCAG: Ensuring Web Accessibility Through Software Testing
Accessibility isn't optional — it's a quality requirement. How to integrate WCAG compliance into your testing workflow.
k6 Performance Testing: A Practical Guide for Product Teams
Load testing doesn't have to be complicated. How we use k6 to catch performance regressions before they hit production — with real examples.
Continuous Testing: None Done
Everyone claims they do continuous testing. Almost nobody actually does. What real continuous testing looks like vs batch QA in disguise.
Prefer a conversation to a blog post?
We will happily spend 30 minutes on your actual problem rather than a general one.