AI coding tools are fast but not always right

Software tools powered by AI are helping researchers update old programs much faster than before. In some cases, these tools speed up the work by up to 60 times. This helps scientists breathe new life into neglected projects.

However, these systems have a major weakness. They sound very smart and confident even when they make mistakes. It is easy for a user to trust the code without checking if the logic holds up.

Now, scientists spend less time writing code and more time double-checking the results. You still need an expert to ensure the science behind the code is sound. Relying solely on AI to handle the math is still too risky.