Loading...
Loading...
Loading...
DYKWYBLSby Island Pitch
Do You Know What You Built Last Summer?!
You vibe coded it. You shipped it. Now let's find out if you actually know what's in there. No shame — just truth.
Do You Know What You Built Last Summer?!
You vibe coded it. You shipped it. Now let's find out if you actually know what's in there. No shame — just truth.
Knowledge Is Protection
Understand the different types of testing and how to build confidence that your app works as expected.
You clicked around your app. The buttons work. The forms submit. The pages load. Ship it, right? On July 19, 2024, CrowdStrike shipped a faulty update to their Falcon Sensor. A wild pointer — an out-of-bounds memory read — that was not caught by testing crashed 8.5 million Windows systems simultaneously. Fortune 500 losses: $5.4 billion. Delta Air Lines filed a $500 million lawsuit. One untested code path. $5.4 billion. That is the distance between "it works on my machine" and "it works."
Testing is writing code that checks whether your other code works correctly. Instead of manually clicking through your app to verify features, you write automated scripts that do it for you — faster, more thoroughly, and every single time you make a change.
There are three main layers. Unit tests check individual functions or components in isolation: does this function calculate tax correctly? Does this component render the right text? Integration tests check that multiple pieces work together: does the API endpoint correctly save data to the database and return the right response? End-to-end (E2E) tests simulate real user behavior: can a user sign up, log in, create a project, and see it on their dashboard?
The AI-generated code you are shipping almost certainly has zero tests. That is not because testing is hard — it is because the AI was optimizing for "make it work right now" rather than "make sure it keeps working." Testing is the difference between a demo and a product.
And testing must include accessibility testing. The WebAIM Million 2024 study found 95.9% of the top million homepages had at least one WCAG error, with an average of 56.8 errors per page. If you are not testing for accessibility, you are not really testing.
Your AI wrote the code. Cool. Now who is going to check the AI's homework?
0 tests detected in most AI-generated codebases. 95.9% of top websites fail WCAG. The testing gap is systemic.
The cost of poor software quality in the US is estimated at $2.41 trillion annually by CISQ, with 2024 estimates reaching $3.1 trillion. Fixing a production bug costs 100 times more than catching it during design.
The CrowdStrike outage is the most dramatic recent example, but the consequences apply at every scale. 40% of companies report at least one critical software failure every quarter. One hour of downtime costs enterprises $300,000 on average. Developers spend 20% of their time fixing bugs — about $20,000 per year in salary cost — and teams spend 30-50% of sprint cycles on defect firefighting instead of building new features.
Your users feel it too. 68% of users abandon an app after just two bugs. 81% of consumers lose trust in brands after major software failures.
And then there is the canonical testing failure: the Therac-25 radiation therapy machine, which killed or severely injured at least six patients between 1985 and 1987 due to a race condition in untested software. The software was designed so it was, in the investigators' words, "realistically impossible to test in a rigorous way." The same bug existed in the previous model, but hardware safety interlocks prevented harm. When those interlocks were removed because developers trusted the software, people died.
Testing is not bureaucracy. It is the difference between a product and a liability.
CrowdStrike crashed 8.5 million computers with one bad pointer. One. Bad. Pointer. And nobody tested for it.
$5.4B CrowdStrike losses. $3.1T annual cost of poor software quality. 100x cost multiplier for production bugs vs. design-phase catches.
Start with the most important functions in your app — anything that handles money, authentication, or data transformation. You do not need 100% coverage to start. Even a handful of tests for critical paths gives you a foundation to build on.
Tests that nobody runs are useless. Set up your CI/CD pipeline to run tests before every deployment. If tests fail, the deploy should stop. This is the single most impactful thing you can do for code quality. CrowdStrike's update went out without adequate automated testing. Do not repeat that pattern.
Your signup flow, login flow, and primary user journey should have end-to-end tests that simulate real user behavior. These catch the bugs that unit tests miss.
With 95.9% of top websites failing WCAG, automated accessibility testing (axe-core, Lighthouse) should run alongside your other tests. Add manual screen reader testing for critical flows. You cannot claim your app works if it does not work for everyone.
Tests that are skipped, ignored, or permanently failing are worse than no tests — they give false confidence. When you change a feature, update its tests. Treat tests as first-class code.
Five items. Start with one unit test for your most important function. Tomorrow, add another. By Friday you have a test suite.
5 items. CI/CD test integration is highest impact. Median setup time: 30 minutes. Prevents every subsequent regression.
Tests are not bureaucracy — they are freedom. The freedom to change code without fear, to ship updates without anxiety, and to sleep at night knowing your app works the way you think it does. CrowdStrike proved that one untested code path can cost $5.4 billion. Therac-25 proved it can cost lives. Start small, automate early, include accessibility, and build the habit. Do Cool Things the Right Way means knowing your stuff actually works — for everyone.
Testing is the difference between 'it works on my machine' and 'it works.' Period.
68% of users abandon after 2 bugs. Automated testing prevents regressions. ROI is immediate and compounding.
Not sure if your project has adequate test coverage? A DYKWYBLS comprehension check examines your codebase and tells you exactly where the testing gaps are — and which ones matter most to fix first. Part of how Island Pitch helps you Do Cool Things the Right Way!
Meet Your Code