What happened: GitHub agent depends on the PRS independently
GitHub has announced a great new capacity on Copilot:
GitHub has revealed an important upgrade to Copilot through Microsoft Build 2025: a cloud -based coding worker capable of drafting and repetitions on cloud requests directly inside GitHub. While previous COPILOT experiments focused on assisting within the editor via VSCODE, this new agent works asymmetically in the cloud, as it benefits from the infrastructure similar to the GitHub procedures.
Developers can now set the tasks directly via GitHub or VsCode:
> @github Open a pull request to refactor this query generator into its own class
COPILOT agent deals with tasks independently: create branches, repetition on PRS based on the comments of reviewing the code, and updates that escalate until the work is accepted – all without touching protected branches. Decally, current CI pipelines, branch protection, and work function remain intact, ensuring “confidence by design”.
With the addition of MCP, developers can even give the agent to up to the tools and external data by directing servers directly in the warehouse settings. This raises us from actual implementation and allows us to focus on describing tasks. This is compatible with the manuscripts agents and others from providing an account for agents for employment and coding.
Also resonated with the broader movement towards cut -off -headed factors – those that work independently, full of tasks, and the report again when the work is ready for review.
Community reactions: awe, caution, fears, and questions.
Personally, the use of the Github Copilot agent was an enjoyable experience when ordering low -level tasks. I passed the moment “Wow” when I took a typical GitHub procedures directly within public relations (or via VsCode chat). He was impressive quickly, and I felt the developer’s experience immediately.
When searching for a community reaction, we found a Copilot coding agent offering a mixture of excitement, curiosity and suspicion of health. The “time saving” was a common praise. Users loved that the public relations project for the agent and the agent allows them to see exactly what was happening, and that they kept choosing when/if the merger is done. Similar to my experience, users shared that they found Copilot to perform a good performance with low -level tasks.
To whom GitHub Lead has answered for Copilot:
Part of society also participated for fear of emptying coding into artificial intelligence, human action may turn into guardianship: offering detailed Jira tickets throughout the day and the symbol created by AI. Users also expressed concern about what this means for beginners. GitHub CEO is deals with this:
In an era in which some claim that more artificial intelligence means lower opportunities for entry level, I think the opposite is true. There was no more exciting time to join our industry.
Based on this concern, security and politics questions were also asked. For example, can the agent be inadvertently displaying sensitive information, such as an inclusion of a secret key in public relations? In addition to the complexity, some developers have indicated that the benefits of using artificial intelligence agents do not come free. You are just really reaping rewards if you are investing the effort in advance; For example, write a good description with clear acceptance criteria.
Enjoy seeing the original development space of Amnesty International (AIIND)? I am helping to build the AIND community, where we cover the most related news with Dev. For busy people, we send one clear news reading 7000+ of artificial intelligence every week – Free and direct to your inbox!
Taking souls: extracting the work of necrosis and developing new habits
GitHub is evolving into the AI improved development platform. It is very similar to the historical growth of the procedure market, we can expect a rise in tools within the Copilot accessories, making Copilot more powerful and covering more aspects of development workflow.
When drawing an analogy of history, this looks closer to the Devops Revolution or the shift to cloud/server computing: worldly infrastructure works are extracted, allowing developers to focus on higher logic. Likewise, the Copilot agent can try a large part of the work of necrosis.
It is also a new skill: demanding and overseeing artificial intelligence in coding. Just as Ci/CD and Cloud have led to new roles, artificial intelligence agents can lead to new roles such as “AI Ornding Engineer”. I think we will see the original startups of artificial intelligence trying more and more cooperation between multiple agents.
One of the cases we believe in is the use of agents who depend on tasks: an agent can create a symbol while another reviews it, or one agent specializing in the front end and the other in the back interface, coordination through issues and PRS. GitHub’s infrastructure can support this type of multi -agent workflow, especially since the MCP allows the various artificial intelligence services series.
From the developer’s point of view, to prepare for this AI’s future future, there are some concrete steps that we recommend when playing with Copilot agent. First, writing important descriptions/issue/public relations will become clear and detailed with admission criteria for a valuable skill. It is a good practice even for human collaborators, but for Amnesty International, it is so very appropriate. This helps to build the muscle of communication with machines about the intention of the code.
Second, invest in the tests. The probability of an artificial intelligence agent depends on the tests to know that it has not broken things is a strong incentive. The highly tested test coverage can become a reliable contributor, while low coverage turns into a dangerous danger. Strengthening your automatic tests and CI pipelines can help develop your projects to benefit more than artificial intelligence.
The final game is to become comfortable to leave Amnesty International a draft thing, and you are going/improving it. In this new era, the developers who are thrived from those who embrace the evolutionary development on specifications, accurately direct artificial intelligence, a symbol of formation, and verify verification with tests.
