Practitioner Newsletter
Deep-dive production content on Agentic DevOps — governed workflows, hookflows, sandboxed tools, and the architecture decisions behind systems that actually run in production.
Cancel anytime · Deep dives, not fluff · Already subscribed? Sign in →
Actual agent definitions, extension code, and pipeline configs from a running 40+ agent system — not sanitized tutorials.
Hookflows, sandboxing, tool surfaces, and structural safety — making the right thing to do the easy thing to do.
Why the system is shaped the way it is — the tradeoffs, the pivots, and the reasoning behind every structural choice.
What broke, what we over-engineered, and what we learned. The stuff nobody else publishes.
Most developers use Copilot for autocomplete. Here's the architecture for building domain-expert custom agents — agents with structured skills, MCP tools, and custom knowledge — that understand your specific workflows at depth.
Most teams trigger AI agents manually. Here's the production architecture that runs 60 autonomous agents on precise schedules — custom cron engine, staggered slots, fresh-agent isolation, and self-healing patterns.
GitOps isn't just K8s deployments. Use Git as a universal control plane for agent behavior, governance, security, and scheduling.
Everyone's building AI agents. Nobody's governing them. Here's the 7-layer governance stack from a production 53-agent system — constitution, approval gates, safety protocols, and isolation patterns.
Every enterprise team is drowning in toolchain sprawl. Here's how to build a real IDP on GitHub using Copilot extensions, IssueOps, hookflows, and golden-path starter repos — with 7 open-source repos you can use today.
Every multi-agent demo shows 3 chatbots in a loop. Here's what happens when you build a 53-agent system that manages a real family — the 4 patterns, the cron architecture, and the orchestration layer that prevents chaos.
How hooks, extensions, and skills give AI agents real power — the complete system powering a 30-extension, 60-skill platform
MCP turns any device into an AI-controllable tool. Here's how I built an 18-tool MCP server that lets GitHub Copilot control my Android phone — and the pattern you can steal for your own hardware.
Visual Studio just productized agent skills — reusable instruction sets for Copilot. I've been running 71 skills across 50 agents for months. Here's the architecture that makes it work at scale.
When AI agents drive your development, instructions aren't enforcement. Here's how I replaced git primitives with governed tools using hookflows and CLI extensions — with the real code from my production system.
Most AI agents forget everything between sessions. Here's the exact 4-tier memory architecture I use across 43 production agents — with real files, real configs, and the skill that enforces it all.
Join practitioners building governed, production-grade agent systems. Get the deep dives every issue.
Subscribe — $19/mo