Adding AI to the development lifecycle has gradually increased the rate of accepted code suggestions and has revolutionized software development.
of research, Issuer Github come after a year in the company co-pilot Coding Assistant is now generally available.of AI Pair Programmer It aims to provide real-time coding suggestions and speed up the development process.
GitHub Copilot has been activated by over 1 million developers and has been adopted by over 20,000 organizations. Over 3 billion lines of generated code.
Research estimates that productivity gains from generative AI coding assistants could boost global GDP by more than $1.5 trillion by 2030. These improvements represent 15 million “efficient developer productivity gains”.
This report is based on an analysis of the coding habits of 934.533 Copilot users, based on code acceptance rates obtained from telemetry.
How many coding suggestions does Copilot accept?
Reports show that an average of 30% of coding suggestions from Copilot are immediately accepted, but that percentage increases as we use it. 28.9% of suggestions were accepted in the first three months of use, increasing to 32.1% in the next three months. The acceptance rate approached 34% at 6 months.
This still means about two-thirds of proposals are rejected, but this represents an improvement over the 26% observed at launch after 12 months of technical preview.
This trend is clear and shows that developers are getting used to the suggestions over time. In addition, programmers gain a deeper understanding of how to use tools to produce meaningful results.
Who is using Copilot?
For this report, developer experience was measured by the number of repository actions on GitHub (before using Copilot). Developers with low repository activity had higher acceptance rates (31.9% on average) compared to the most experienced developers (26.2% on average).
Related resources
AI Inference with AMD EPYC™ Processors
Provides an excellent platform for CPU-based AI inference
Download for free
Inexperienced developers therefore get the most obvious productivity benefits from AI coding assistants.
GitHub will continue to evolve as tools evolve (e.g. Implementing a natural language model), about 80% of the code will be written in AI. Research describes this as the democratization of software development, as it allows people with no technical background to build applications.
In a recent discussion at Kubecon, Cloud Native Computing Foundation CTO Chris Aniszczyk said that Copilot is “probably 80% achievable” for the basics at this point, but also said it’s “not perfect.” added.he also pointed out that Ongoing issues with code attribution.
What does Copilot mean for junior developers?
As the rise of AI pair programming tools continues, the skills of young developers must also change. The study compares technology change with moving from punch cards to compilers and high-level programming languages.
It also suggests the potential for new business models compared to the advent of steam engines and the internet and the higher order jobs that are being generated.
Even though the boom seen in recent years has subsided, the demand for developers remains strong. claims to be.
Another interpretation of the trends shown in this study is that productivity gains come at the expense of junior developers, effectively closing the pipeline for senior staff.
GitHub disagrees, stating, “This clash between AI and software developers will not lead to fewer jobs for developers. AI will enable developers to do more and accelerate human progress.” .”
