Recently there was a story about an updater causing a $8,000 bill because there was a lack of basic automated tests to catch the issue. [0]<p>The big lesson here is that you should actually test the code you write and also write automated tests to check any code generated by an LLM that the code is correct in what it does.<p>It is also useless to ask another AI to check for mistakes created by another LLM. As you can see in the post, both of them failed to catch the issue.<p>This why I don't take this hype around 'vibe-coding' seriously since not only it isn't software engineering, it promotes low quality and carelessness over basic testing and dismisses in checking that the software / script works as expected.<p>Turning $70 problems found in development into $700,000+ costs in production.<p>There are no more excuses in not adding tests.<p>[0] <a href="https://news.ycombinator.com/item?id=43829006">https://news.ycombinator.com/item?id=43829006</a>