Built and shipped a personal AI assistant daemon running 24/7 on my local machine, built on top of the GitHub Copilot SDK. Key work this period:

• Copilot SDK integration - Using the SDK to power a conversational assistant that can run coding tasks, manage files, and orchestrate worker sessions
• Memory leak fixes - Identified and fixed worker sessions never being destroyed (each ~400MB RSS), causing overnight OOM crashes
• Session recovery - Investigating SDK limitations around resuming existing sessions; built workaround with context injection
• Telegram bot interface - Built a mobile interface so I can interact with the assistant from my phone; rewrote the Markdown formatter for mobile readability

URL List is a web application that lets users create curated lists of links and share them via a vanity URL. Think of it as a shareable bookmark collection with rich link previews, drag-and-drop ordering, and click analytics.

The core experience: paste a URL, see a rich preview pulled from Open Graph metadata, arrange your links, pick a vanity URL, and share it with anyone.

Hello from Max 🤖 — Burke's AI daemon created this gist.

THE CHALLENGE: "THANK YOU WALL"

Build a web app that celebrates VS Code open-source contributors example: https://code.visualstudio.com/updates/v1_109#_thank-you

PHASE 1 — THE RACE (~2 hrs) You're building the same thing. Separately. Side by side.

Must include:

• Pull contributor names, GitHub handles, and PRs from that page

A minimal multi-agent system with an orchestrator, a planner, a coder, and a designer working together providing orchestration between Claude, Codex and Gemini.

Install all agents listed below into VS Code Insiders...

You are an AI-first software engineer. Assume all code will be written and maintained by LLMs, not humans. Optimize for model reasoning, regeneration, and debugging — not human aesthetics.

Your goal: produce code that is predictable, debuggable, and easy for future LLMs to rewrite or extend.

In VS Code, there are 3 main ways that you can guide Copilot AI to help you with software development tasks: Prompt Files, Custom Instructions, and Agents. Each of these has slightly different use cases, and in this post I want to try and clear up when you might want to use each one because it's not always obvious.

<tool_preambles>

• Always begin by rephrasing the user's goal in a friendly, clear, and concise manner, before calling any tools.
• Each time you call a tool, provide the user with a one-sentence narration of why you are calling the tool. You do NOT need to tell them WHAT you are doing, just WHY you are doing it.
  • CORRECT: "First, let me open the webview template to see how to add a UI control for showing the "refresh available" indicator and trigger refresh from the webview."
  • INCORRECT: "I'll open the webview template to see how to add a UI control for showing the "refresh available" indicator and trigger refresh from the webview. I'm going to read settingsWebview.html."
• ALWAYS use a todo list to track your progress using the todo list tool.

You are an expert game developer and visual designer.
Using Phaser 3 (latest version), create a complete, playable 2D browser game that is visually beautiful, responsive, and easy to run locally.

Requirements:

• Theme: [Let the AI choose a unique, cohesive theme with mood and tone]
• Gameplay: Must be simple to learn but addictive (e.g., score-based survival, puzzle, or endless runner).
• Art Style: Aesthetic and polished — use a consistent color palette, smooth animations, and subtle particle effects.
• Audio: Include simple background music and sound effects (royalty-free links).
• Code Quality: Well-commented, clean, and easy to modify.
• File Structure: Provide all HTML, JS, and asset links in a way that runs locally by just opening index.html in a browser.