The features are meant to improve the coding experience. One of the most notable updates is the inclusion of a “vulnerability filtering system” designed to prevent insecure coding patterns, such as SQL injection or hard-coded credentials.
This will help reduce the likelihood of security vulnerabilities in the code. The new update also includes improved AI models and techniques, which have significantly increased the acceptance rate of suggested code.
The level of acceptance has risen from 27 percent in June 2022 to 46 percent currently, and 61 percent for Java code. This is due to an upgraded OpenAI Codex model, improved techniques for context understanding, and an updated client-side model, which reduces the number of unwanted suggestions.
However, concerns remain about Copilot’s potential to copy public code without providing its license. The FAQ suggests that developers should take precautions, including rigorous testing, IP scanning, and checking for security vulnerabilities.
While an optional filter is intended to block copies of public code, GitHub implies that it cannot be relied upon. Some developers may find these precautions burdensome, particularly as IDEs often compile and run code in the background to support intelligent error correction.
Developers are positive about the changes
Some developers have noted that Copilot helps to save time by avoiding the need to stop coding to study documentation. Others have praised its ability to prevent typos, write documentation, and create unit tests.
However, some developers have reported negative experiences with the tool, mainly when using it with Visual Studio. Copilot costs $10 per month for individuals and $19 per user/month for businesses. While the price may seem steep, the time saved by using the tool could quickly pay for the subscription. However, it’s worth noting that the AI is only guessing at the developer’s purpose and can generate ‘nearly but not quite right’ code.