Blog

Ideas worth proving

Plain-spoken, evidence-led writing on QA automation, AI, data science, and engineering. The biggest pain points, and what actually solves them.

EnergyQA AutomationEngineeringAI & AutomationData Science
Energy8 min read

Automating curtailment claims: the revenue renewable producers leave on the table

Grid-instructed curtailment costs renewable energy producers worldwide material revenue, and the manual, spreadsheet-based claim process loses even more. Here's how automation turns curtailment claims into a fast, accurate, fully-auditable workflow.

Nullius Technologies
QA Automation8 min read

Testing payment and SWIFT flows: where SA banking QA gets hard

What makes payment, settlement, and SWIFT message testing uniquely demanding in South African banking, and how to build automated coverage that regulators and reconciliation teams can trust.

Nullius Technologies
QA Automation6 min read

Buying a test suite is easy. Owning it is the hard part

Most test automation projects decay because the team that inherits them was never trained to own them. Here's how structured upskilling keeps a suite alive long after handover.

Nullius Technologies
QA Automation7 min read

CI/CD test gates that catch bugs without slowing releases

How to design CI/CD test gates that stop defects reaching production without turning your pipeline into a bottleneck that engineers learn to route around.

Nullius Technologies
QA Automation6 min read

Why SA banks are saving millions by ditching manual regression

Manual regression testing quietly costs South African enterprises millions a year. Here's how test automation pays for itself, with the numbers.

Nullius Technologies
Engineering4 min read

Nullius in Verba: the principle behind our name

Where the name Nullius Technologies comes from, and why "take nobody's word for it" is the right foundation for QA, data science, and engineering.

Nullius Technologies
QA Automation7 min read

Playwright vs Selenium in 2026: what SA teams should actually use

A practical, no-hype comparison of Playwright and Selenium for South African engineering teams choosing a test automation framework in 2026.

Nullius Technologies
AI & Automation7 min read

Giving AI a consistent voice: why generic output kills adoption

Why AI-generated content so often sounds generic, what it takes to make a model write in a specific person's or brand's voice, and why that capability drives real adoption.

Nullius Technologies
Engineering8 min read

Building a full AI product, not another impressive demo

What it actually takes to ship a complete, production-ready AI product, from capture to output to publishing, using a real end-to-end build as the example.

Nullius Technologies
Engineering7 min read

Web performance isn't vanity: a fast site makes money

Why web performance drives conversion and SEO, what Core Web Vitals and a Lighthouse score actually measure, and how to build fast sites without sacrificing features.

Nullius Technologies
Data Science6 min read

When not to use machine learning

Machine learning is powerful and frequently the wrong tool. An honest guide to when a model earns its complexity and when a rule, a query, or a dashboard does the job better.

Nullius Technologies
Engineering7 min read

Your most important process shouldn't run on a spreadsheet

When a critical process depends on a fragile spreadsheet and the person who built it, you're carrying real operational risk. How to tell when it's time to build a proper tool, and how to do it without a huge project.

Nullius Technologies
QA Automation7 min read

Load testing: find your breaking point before your users do

What load and performance testing actually tells you, how to run it meaningfully, and why finding your system's limits deliberately is far cheaper than finding them by accident.

Nullius Technologies
Data Science8 min read

Catching fraud and anomalies with machine learning

How machine learning anomaly and fraud detection works, where it beats rules, where it doesn't, and how to deploy it without drowning your investigators in false positives.

Nullius Technologies
QA Automation7 min read

Mobile test automation without the pain

Why mobile test automation is harder than web, and a practical approach using Appium and device strategy to get reliable coverage across the devices that actually matter.

Nullius Technologies
Data Science9 min read

Forecasting transaction volumes: lessons from a top-3 national finish

The techniques behind a 3rd-of-251 finish in Nedbank's ML challenge: temporal panels, leave-one-cutoff-out target encoding, model diversity, and segmented blending, explained for teams forecasting skewed count data.

Nullius Technologies
Engineering7 min read

Shipping an MVP that doesn't become next year's technical debt

How to build a minimum viable product that validates your idea quickly and survives contact with success, instead of becoming the technical debt that slows you down for years.

Nullius Technologies
AI & Automation8 min read

Keeping LLMs reliable in production: guardrails over hope

How to move an LLM feature from an impressive demo to a dependable production system, with evaluation, guardrails, structured outputs, and human-in-the-loop review.

Nullius Technologies
Data Science7 min read

Predicting customer churn before it happens

How a churn prediction model works, what data it needs, and how to turn its output into retention actions that actually move the number, with an honest look at where it adds value and where it doesn't.

Nullius Technologies
QA Automation7 min read

API testing: the highest-leverage tests you can write

Why API tests give you more coverage per unit of effort than UI tests, what to test at the API level, and how it fits into a balanced automation strategy.

Nullius Technologies
AI & Automation8 min read

Document processing automation that actually works in production

Document extraction looks easy in a demo and fails in production without the right guardrails. How to build invoice, form, and statement processing that stays accurate at scale.

Nullius Technologies
QA Automation7 min read

Why your tests are flaky, and how to actually fix them

The real root causes of flaky automated tests, bad waits, shared state, wrong test level, brittle selectors, and the concrete fixes that make a suite trustworthy again.

Nullius Technologies
Data Science7 min read

The data you already have is worth more than you think

You don't need a data lake or a data science team to get value from your data. How to go from scattered, messy data to a working prototype that answers a real business question.

Nullius Technologies
AI & Automation7 min read

n8n: the automation backbone most teams overlook

Why n8n has become our default for workflow automation, how it compares to RPA and no-code SaaS tools, and where it fits in an AI-powered automation stack.

Nullius Technologies
Engineering6 min read

Testing isn't a phase, it's a culture

Why treating testing as a final phase guarantees expensive, late-caught defects, and how the best teams build quality into the whole delivery process instead.

Nullius Technologies
AI & Automation7 min read

AI automation ROI: how to pick what to automate first

A practical framework for choosing which workflows to automate first, so your AI and automation budget pays back in weeks, not quarters.

Nullius Technologies