Blog

How my website keeps growing without me opening the repo

Routing, fusion and subagents get used interchangeably in almost every multi-agent write-up. They solve different problems at different moments. Twelve orchestration patterns, one diagram each, a tool table, and when the extra complexity actually earns its keep.

A hands-on deep dive into the bounded agent jury: freeze a target, hand it to blind read-only judges, classify agreement as confirmed, suspect or contradiction, and know when it actually needs different models instead of just fresh context.

Orca runs Claude Code, Codex and OpenCode in isolated git worktrees. Its orchestration skill adds the missing layer: one coordinator agent that dispatches work, waits for results, and pings your phone when it needs a decision.

A practical, framework-agnostic guide to agentic design patterns - prompt chaining, routing, parallelization, orchestrator–workers, evaluator–optimizer, and ReAct grounded in Anthropic's Building Effective Agents, with when-to-use rules and diagrams.

A practical guide to conversation compaction for AI agents: trimming, rolling summaries, structured state, retrieval, prompt compression, and the failure modes that quietly break long-running tasks.

Andrej Karpathy's four principles for AGENTS.md — Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution — became a 179k-star GitHub repo. Here's what they are and why they work.

Git worktrees let you check out multiple branches simultaneously in separate directories. For AI agents running in parallel, they're not a nice-to-have — they're the only way to avoid chaos.

Harness engineering is the complete environment you build around an AI model — tools, constraints, feedback loops, and context. Learn what it is, how it differs from prompt and context engineering, and how to start building one.

Most teams are not bad at giving AI context. They're bad at putting the right context in the right layer. Here's a practical, architecture-aware framework for AGENTS.md, repo-wide instructions, and path-specific overrides.

Learn how VS Code Copilot hooks let you run shell commands at key agent lifecycle points — block dangerous operations, auto-format code, inject context, and build audit trails.

Learn how to use WebMCP to expose structured tools from your website to in-browser AI agents. Set up Chrome Canary, implement declarative and imperative APIs, and build agent-aware forms — no MCP server required.

Deep dive into Agent Skills best practices, real-world patterns, editor conventions, npm-like installation workflows, and security considerations for building production-ready AI workflows.

Learn the fundamentals of Agent Skills — how SKILL.md files work, the progressive disclosure model, and when to use skills vs. rules for your AI coding assistant.

Learn how to create and execute custom prompts in VS Code using the .github/prompts folder for faster, more consistent AI interactions.

Learn how to create AI agents that understand your entire codebase using Repomix and GitHub Copilot. Build intelligent assistants for your development team.

Master GitHub Copilot with strategic prompting and context management. Learn how to guide AI assistance effectively and make Copilot work for you and not against you.

Avoid these common pitfalls when using GitHub Copilot. Learn how to write better prompts, review AI suggestions critically, and iterate effectively for better code quality.