GitHub Engineering
https://github.blog/Why it matters: Custom agents in GitHub Copilot empower engineering teams to embed their unique rules and workflows directly into their AI assistant. This streamlines development, ensures consistency across the SDLC, and automates complex tasks, boosting efficiency and adherence to standards.
- •GitHub Copilot now supports custom agents, extending its AI assistance across the entire software development lifecycle, not just code generation.
- •These Markdown-defined agents act as domain experts, integrating team-specific rules, tools, and workflows for areas like observability, security, and IaC.
- •Custom agents can be deployed at repository, organization, or enterprise levels and are accessible via Copilot CLI, VS Code Chat, and github.com.
- •They enable engineers to enforce standards, automate multi-step tasks, and integrate third-party tools directly within their development environment.
- •A growing ecosystem of partner-built agents is available for various domains, including security, databases, DevOps, and incident management.
Why it matters: This article highlights the engineering complexities and architectural decisions behind building a robust, local-first distributed system for the physical world. It showcases how open-source governance can be a technical requirement for long-term project integrity and user control.
- •Home Assistant is a fast-growing open-source home automation platform, used in over 2 million households and attracting 21,000 contributors annually.
- •It champions a local-first architecture for privacy and interoperability, enabling control of thousands of devices on user hardware without cloud dependency.
- •The platform abstracts diverse devices into local entities with states and events, acting as a distributed event-driven runtime for complex home automations.
- •This local-first approach presents significant engineering challenges, demanding optimizations for device discovery, state management, and network communication on constrained hardware.
- •Governance by the Open Home Foundation ensures its open-source integrity, protecting against commercial acquisition and maintaining its core local-first philosophy.
Why it matters: This tool enhances developer productivity by enabling parallel execution and orchestration of AI coding agents, centralizing task management and review. It shifts the mental model from sequential to concurrent work, optimizing development workflows.
- •GitHub's new Agent HQ mission control provides a unified interface for managing Copilot coding agent tasks across multiple repositories.
- •The tool facilitates a shift from sequential to parallel task execution, allowing engineers to assign and orchestrate multiple agent tasks concurrently.
- •Effective orchestration involves crafting clear, contextual prompts and leveraging custom agents for consistent results.
- •Engineers must actively monitor agents for signals like failing tests, scope creep, or misinterpretation, intervening with specific guidance when necessary.
- •While parallel processing is ideal for research, analysis, documentation, and security reviews, sequential workflows remain suitable for dependent or complex tasks.
- •Mission control centralizes assignment, oversight, and review, streamlining the development workflow and enhancing productivity.
Why it matters: This guide helps individuals find practical and fun GitHub-themed gifts for developers, enhancing their daily work and personal life with branded merchandise. It's relevant for celebrating developer culture and community.
- •The article presents a holiday gift guide featuring a range of GitHub-branded merchandise for developers.
- •It suggests festive apparel such as ugly holiday socks, beanies, and sweaters, some available in a Black Friday sale.
- •Hydration and coffee solutions are highlighted, including various bottles, mugs, and tumblers for different settings.
- •Novelty items like the GitHub Copilot Amazeball are offered for fun and decision-making.
- •Workspace enhancements include custom key caps, a recycled desk mat, and a MiiR laptop backpack.
- •The guide also features youth apparel, ensuring gifts for younger developer enthusiasts.
Why it matters: This article highlights Python's enduring appeal, its foundational design principles emphasizing readability and accessibility, and its continued dominance in AI and data science, offering insights into language evolution and developer preferences.
- •Python, created by Guido van Rossum, emerged to simplify programming by offering a safer, more expressive alternative to C and shell scripting.
- •Despite TypeScript's recent lead on GitHub, Python grew 49% in 2025, maintaining its status as the default language for AI, science, and education.
- •Its core design emphasizes readability, intuitive syntax, friendly error messages, and a rich standard library, fostering accessibility.
- •Python's open-source nature, cross-platform support, and strong community are key to its versatility and widespread adoption.
- •The language's "irreverent" name reflects a deliberate choice to make programming less intimidating and more welcoming.
Why it matters: This article provides essential security principles for developing and deploying AI agents, addressing critical risks like data exfiltration and prompt injection. It offers practical guidelines for ensuring human oversight and accountability in agentic systems.
- •GitHub employs agentic security principles for AI agents like Copilot, balancing usability with security through a human-in-the-loop design.
- •Key risks for agentic AI include data exfiltration, impersonation/action attribution, and prompt injection.
- •Security controls ensure all context is visible, agents are firewalled, and access to sensitive data is limited.
- •Agents are prevented from making irreversible state changes without human approval, such as creating pull requests instead of direct commits.
- •Actions are clearly attributed to both the initiating user and the agent, ensuring accountability.
- •Context gathering is restricted to authorized users with appropriate repository permissions.
Why it matters: These proposed patent rule changes could significantly increase legal risks and costs for developers and startups, hindering innovation and open-source projects. It makes challenging bad patents much harder, impacting the entire tech ecosystem.
- •USPTO's new rules propose to significantly restrict Inter Partes Review (IPR), making it harder to challenge low-quality patents.
- •IPRs were created to provide an efficient, affordable way for developers and startups to challenge questionable patents, supporting innovation.
- •The 2025 proposal introduces strict rules blocking IPR petitions in common scenarios, unlike prior less restrictive proposals.
- •It would prevent developers from challenging patents if another party previously failed, and require waiving invalidity defenses in court.
- •These changes escalate litigation risks and costs for developers, startups, and open-source projects, impeding open innovation.
- •The article urges developers to file comments against these proposed rules to protect innovation.
Why it matters: This article details advanced techniques in training AI for developer tools, showcasing how custom data collection, SFT, and RL overcome challenges in real-time code prediction. It's crucial for engineers building AI-powered developer experiences and understanding practical LLM deployment.
- •GitHub Copilot's Next Edit Suggestions (NES) uses a custom, low-latency model designed to predict developers' next code edits in real time.
- •Initial attempts with general LLMs and pull request data failed; a custom, high-quality dataset derived from real-time editing sessions was crucial for training.
- •The foundational NES model was developed using Supervised Fine-Tuning (SFT) on this specialized dataset.
- •Reinforcement Learning (RL), incorporating a custom 'grader' model, further refined the NES model, addressing SFT limitations by leveraging unlabeled data and explicitly defining criteria for 'bad' suggestions.
- •This 'AI-native' approach emphasizes end-to-end co-design of model training, prompt engineering, and user experience for seamless IDE integration.
- •Recent improvements focus on prompt optimization to reduce latency and enhance the relevance and quality of suggestions.
Why it matters: Optimizing tool selection for LLM agents significantly boosts performance and reliability. This approach reduces latency and improves success rates for AI assistants like GitHub Copilot, making them faster and more effective for developers.
- •GitHub Copilot Chat's performance was hindered by reasoning across hundreds of tools via the Model Context Protocol (MCP).
- •New systems, embedding-guided tool routing and adaptive tool clustering, were developed to optimize tool selection for LLM agents.
- •The default toolset was reduced from 40 to 13 core tools, and 'virtual tools' were introduced to functionally group similar tools.
- •Adaptive tool clustering uses Copilot's internal embedding model and cosine similarity to efficiently group tools, replacing slower LLM-based categorization.
- •Embedding-guided tool routing pre-selects the most semantically relevant tools based on query embeddings, reducing unnecessary exploratory calls.
- •These optimizations improved success rates by 2-5 percentage points on benchmarks and reduced response latency by an average of 400 milliseconds.
Why it matters: This article provides actionable insights for developers to leverage GitHub Copilot's custom agents effectively. By following these best practices, engineers can create highly specialized AI assistants that improve productivity and code quality, streamlining development workflows.
- •GitHub Copilot's new `agents.md` feature enables custom, specialized AI assistants for tasks like documentation or testing.
- •Successful `agents.md` files are highly specific, defining clear personas, executable commands, explicit boundaries, and concrete code examples.
- •Analysis of over 2,500 repositories reveals best practices: place commands early, use code snippets for style, set strict "never do" rules, and detail the exact tech stack.
- •Effective agent definitions should cover six core areas: commands, testing, project structure, code style, Git workflow, and boundaries.
- •Begin by creating agents for simple, specific tasks (e.g., linting, doc generation) with a clear name, description, and persona.
- •Copilot can help generate initial `agent.md` files, which users then customize with project-specific details and YAML frontmatter.