/writing
From discovery to framework to failure log to thesis. These trace the arc from infrastructure engineer to context orchestration — documented as it happened.
Convergent Architecture: Three Domains, One Principle
→Infrastructure, software architecture, and agent orchestration all arrive at the same answer independently. The intelligence lives in accumulated context, not individual execution.
AgentOps: It's Not Agent Orchestration. It's Context Orchestration.
→Everyone, including me, has been calling it agent orchestration. After months of shipping code with AI agents, I realized we've been naming the wrong thing. Here's what context orchestration actually looks like.
The Knowledge Flywheel: How AI Teams Compound Intelligence
→Every AI session generates learnings. Most teams throw them away. The flywheel captures, promotes, and re-injects knowledge so each session starts smarter than the last.
Multi-Agent Orchestration: Lessons from Running 31 Repos with AI
→7,413 commits across 31 repos. What actually works when you coordinate multiple AI agents on real infrastructure, and what breaks spectacularly.
Shadow Integration: What Gaming Taught Me About Learning to Code
→Subnets were impossible until they weren't. Then networking. Then Kubernetes. Then Python. Each was a boss fight I wasn't supposed to win.
Spec-Driven Development: Why Your AI Agent Needs a Blueprint
→Writing specs before letting agents code is shift-left for AI development. The spec is the leverage point. Get it right and the implementation is mechanical.
The Validation Bottleneck: Why AI Output Quality Is the New CI/CD
→AI generates code in seconds. Validating it takes minutes. The bottleneck isn't generation anymore. It's knowing whether the output is any good.
Devlog #5: When the Platform Catches Up
→Claude 4.6 ships native multi-agent. I tried it. The plumbing is great. The orchestration is still ours.
Why ADHD Brains Are Built for AI Coding
→Traditional coding has a feedback loop problem. ADHD brains check out when the loop is too long. AI coding compresses it to 10-30 seconds. This isn't a workaround. It's an architectural match.
Devlog #4: Why Your AI Coding Specs Keep Failing
→The model improves every quarter without you. Your spec-writing skill? That's what actually matters.
Devlog #3: From Vibe Coder to Vibe Engineer
→The meta keeps shifting. Understanding the system, not the builds, is what transfers.
The REPL Is Dead. Long Live the Factory.
→2026 is the year of the orchestrator. The REPL pattern isn't dying. It's evolving. Here's what it becomes.
Devlog #2: How to Actually Use Your .claude/ Directory
→The practical guide to commands, skills, and agents - and why you probably have too many.
Devlog #1: What Running 8 Parallel AI Agents Taught Me
→Why AI coding tools break past a single session, and the architecture shift that fixed it.
The Gutenberg Moment for Code
→The printing press took 77 years to spark the Reformation. LLMs are moving 100x faster. The best time to learn was 2024.
Getting Started with Vibe Coding
→A practical guide to calibrating AI trust. Install vibe-check, pick a task, declare a level, and measure the results.
12-Factor AgentOps
→Shift-left validation for coding agents. DevOps principles applied to AI-assisted development.
From WoW Raids to AI Agents
→Gaming skills transfer directly to AI coding. Here's how to recognize which ones you already have.
Building This Website with Vibe-Coding
→First-time web dev: Next.js, Vercel, OpenGraph. Built with Claude Code at L3-L4.
Building vibe-check
→Measure your AI collaboration with 5 core metrics from git history.