Gradle 8.0 Provides Improved Kotlin DSL and Build Times

Published on march 2, 2023

Also, GitHub has begun using artificial intelligence to detect vulnerabilities in Copilot suggestions by blocking insecure coding patterns in real time.

GitHub reports that three major technical improvements have been made to Copilot, beginning with the adoption of a new OpenAI Codex model for synthesis of better code.

According to GitHub, FIM delivers better results more consistently and without adding latency.

Additionally, GitHub has improved the Copilot extension for Visual Studio Code to reduce the frequency of unwanted suggestions, which can disrupt the development process. Consequently, Copilot now considers a number of factors related to the user’s context, including whether the last suggestion was accepted. GitHub’s own metrics indicate that this new approach resulted in a 4.5% reduction in unwanted suggestions.

The cumulative effect of all these changes, as well as those released previously, has been an increase in the overall acceptance rate for Copilot code suggestions, which has grown from 27% in June 2022 to 35% in December 2022.

Another area where GitHub has begun to apply AI is the prevention of vulnerabilities in code generated by Copilot. This is accomplished by identifying insecure coding patterns, such as hardcoded credentials, SQL injection, and path injection. In the event that an insecure pattern is identified, it is blocked and a new suggestion is generated.

According to GitHub, the LLM will be further enhanced to be able to identify vulnerable code and distinguish it from secure code in order to radically improve how developers avoid introducing vulnerabilities into their code.

Both the new AI model and vulnerability filtering system are available in GitHub Copilot for Individuals and Copilot for Business.