A study published in IEEE Transactions on Software Engineering evaluated OpenAI's ChatGPT for coding capabilities. The AI showed varied success rates, from 0.66% to 89% depending on factors like task difficulty and the programming language. ChatGPT excelled at solving older problems but struggled with newer ones due to training dataset limitations. The study also highlighted ChatGPT's ability to provide code with less runtime and memory overhead than human solutions but noted security concerns and the need for developers to guide the AI to avoid vulnerabilities.