LLMs have lowered the bar for the unskilled person to create shit software. I have used Opus 4.6 on a number of projects, and it still spits out buggy, and sometimes, flat out broken code. I was actually surprised when it completely hallucinated the names of query params for an HTTP request in my code, when in the prompt I had explicitly given it the exact names it needed to use. I thought these frontier models were supposed to be game changing.

The query parameter issue is a pattern I see a lot. The model has thousands of examples of "how HTTP requests usually look" from its training process. When your input data conflicts with the pattern, the training data takes precedence.

Interestingly, the model doesn't "know" that it's ignoring you. From its perspective, it has retrieved a "meaningful" pattern—virtual parameter names that probably fit common conventions it saw during training. Your actual request simply... wasn't documented.

> LLMs have lowered the bar for the unskilled person to create shit software.

So? Demand the source code. Run your own AI to review the quality of the code base. The contracting company doesn't want to do it? Fine, find one that will.