A static Astro website for a South Wales carpentry and construction business, built with no CMS, database, or client-side framework, and deployed to GitHub Pages on a custom domain.

One model reviewing its own plan tends to agree with itself. So I built a cleanup workflow shaped like a jury: separate critics from a deliberate mix of model families weigh the evidence, with a simple forcing function that keeps neighbouring reviewers off the same family. The orchestrator presides as judge, refuting rather than merging, votes promote shared findings, and a post-approval diff check can overrule a verdict that was unanimous and still wrong.

Building with one agent is familiar. Orchestrating a fleet of them in parallel across local, background, and cloud surfaces is where it gets real. This session covers the decisions that matter, how to decompose work across agents, when to fork vs. delegate, and how to verify quality when agents outnumber you — with patterns for multi-agent workflows you can apply to your own codebase.

An interactive mental model for agentic AI that explains how instructions, retrieved context, session state, memory, MCP servers, sandbox execution, tool results, and the Think-Act-Observe loop fit together.

Conversation state and retrieved context lead naturally into memory, but only if we're clear about what memory is for. Rules, skills, and instruction files package what you already know. Memory should capture what the work itself teaches the system, and that means reflection, verification, and forgetting matter just as much as recall.

Context engineering gets more useful when the knowledge it depends on is packaged for reuse. In this post I map the portable core (AGENTS.md and SKILL.md), Copilot-specific concepts like custom instructions, agents and prompt files, and how to decide what knowledge belongs where.

One model reviewing its own plan tends to agree with itself. So I built a cleanup workflow shaped like a jury: separate critics from a deliberate mix of model families weigh the evidence, with a simple forcing function that keeps neighbouring reviewers off the same family. The orchestrator presides as judge, refuting rather than merging, votes promote shared findings, and a post-approval diff check can overrule a verdict that was unanimous and still wrong.

Conversation state and retrieved context lead naturally into memory, but only if we're clear about what memory is for. Rules, skills, and instruction files package what you already know. Memory should capture what the work itself teaches the system, and that means reflection, verification, and forgetting matter just as much as recall.

Context engineering gets more useful when the knowledge it depends on is packaged for reuse. In this post I map the portable core (AGENTS.md and SKILL.md), Copilot-specific concepts like custom instructions, agents and prompt files, and how to decide what knowledge belongs where.

Better prompts help, but they're only part of the story. Context engineering is the craft of designing what an AI agent sees, when it sees it, and how that changes across the session. The goal isn't a bigger context window. It's a more effective one.

There's a pattern I've seen play out across dozens of DevRel conversations, confirmed in my MBA dissertation research: teams without a clear golden thread from company strategy to daily activity get 'randomised' by whoever asks most urgently. Here's what the research says about why it happens and how to build your way out of it.

DevRel is often framed as the voice of the developer. My research suggests a broader job: gathering representative feedback, reducing friction, and showing developers what changed.

Building with one agent is familiar. Orchestrating a fleet of them in parallel across local, background, and cloud surfaces is where it gets real. This session covers the decisions that matter, how to decompose work across agents, when to fork vs. delegate, and how to verify quality when agents outnumber you — with patterns for multi-agent workflows you can apply to your own codebase.

Join VS Code, Live! at Microsoft Build. For the first time ever, VS Code, Live! is going on stage—in person—and you’re invited! Get an inside look at what the team’s building, featuring live conversations with devs and creators. Whether you’re deep into VS Code or just curious about what’s next, this is your dev-first, code-forward pass to what’s new.

Whether you’re addressing an issue, iterating on a pull request, or bootstrapping a project, GitHub Copilot Workspace helps jumpstart your tasks by describing what you want in natural language. You remain in control as you move between tasks, specs, plans, and code. Join GitHub's Chris Reddington, senior program manager of DevRel, and Cole Bemis, research engineer on GitHub Next, for an introduction to Copilot Workspace, a Copilot-native dev environment launched in April 2024 by GitHub Next. Learn how Copilot Workspace works, how we got here, and what we've learned so far from the technical preview.

Come to this session to be the first in the world to get a deep-dive demo of the exciting Claude 3.5 Sonnet model from Anthropic built right into GitHub Copilot. Learn what this powerful new model will help you achieve and why you should use it for your development teams.

As AI continues to reshape the way businesses innovate, the conversation shifts beyond the mere integration of new technology. Chris emphasizes that adopting AI-powered software development is a significant cultural shift, requiring more than just technical tools. It demands a top-down approach, where leadership support and change management are critical to fostering a progressive culture within development teams.

Ready to explore our blueprint for AI-powered development? Join us to network with local tech leaders and learn the key to addressing tech debt, modernizing the software development lifecycle, and transforming your enterprise.

Chris Reddington demonstrates "Flight School," a custom Next.js application built to personalize his learning journey using the GitHub Copilot SDK. See how he leverages agentic workflows to generate daily coding challenges based on his GitHub profile, evaluate solutions against test cases, and automatically export projects to new repositories.

Join us for Rubber Duck Thursdays! A lighthearted and informal stream where we live code on some projects. This week we explore how to bring the power of GitHub Copilot into your apps with the GitHub Copilot SDK, building hands-on examples and discussing patterns for integrating AI-powered coding assistance directly into developer tools and workflows.

Chris showcases Timestamp, an open-source countdown timer app built with vanilla TypeScript featuring an extensible theme system, world map wall clock, and 2500+ unit tests with 300+ Playwright end-to-end tests. He deep-dives into structuring GitHub Copilot custom instructions vs agents vs prompt files, demonstrates manager-specialist agent patterns, and walks through GitHub Actions automation for issue forms and theme scaffolding.

Chris builds a New Year countdown app from scratch using Vite, TypeScript, and Tailwind CSS, leveraging GitHub Copilot's Plan agent for iterative requirements and a custom UI Performance Specialist agent. The stream covers adding time zone support with a world map, a fireworks celebration theme using fireworks.js, and a test-driven development approach.

Chris reviews the latest GitHub changelog including CodeQL updates, GPT-5.1 Codex Max preview, GitHub MCP server tool-specific configuration, and Enterprise Server 3.19 GA. He then explores the major VS Code December release featuring Agent HQ, background agents with git worktrees, sub-agents, and custom agents for organizations, while continuing to build custom agents for a game MCP server project.

Chris continues building custom agents for the software development life cycle. After reviewing the changelog including Copilot Spaces updates and assigning issues to Copilot via API, he explores the awesome-copilot repository for agent inspiration, brainstorms SDLC agent ideas like test writers and security reviewers, and live-builds a GitHub Actions workflow agent with a companion instructions file.