What’s your vibe?

Can you spot an issue in "這塊士很肥"? What about this code snippet: funcs = [lambda: i * 2 for i in range(5)]? They are subtle and don't jump out immediately – the third character is miswritten. Also, the lambda function should be written as lambda x=i: x * 2 due to Python’s late binding behavior.

These examples might seem contrived, but they illustrate an important point. LLMs handle language and code well, but occasionally make mistakes, and typically, it requires a trained eye to catch these errors. This makes LLMs invaluable tools for those who are already skilled in their craft.

The challenge is with beginners. We are all novices in many areas, and there’s nothing wrong with that. But just as I wouldn’t try to edit a Japanese newspaper without knowledge of Japanese, you probably shouldn’t base significant financial decisions on an ML model you vibe coded.