Blog
-
2026-03-24
The Work Between Systems
Most enterprise software captures what happened. It doesn’t capture the reasoning that connects the dots. That gap — the in-between work — is what AI can finally start to close.
-
2026-02-18
Git as the Shared Brain: My Hybrid Agentic Dev Workflow
Running two AI coding agents across phone and terminal without losing context, and why Git is a better coordination layer than conversation history.
-
2026-02-16
VSCode Was Built for the Agentic Age Before It Existed
The architectural decisions Microsoft made in VSCode between 2016 and 2022 accidentally created the perfect foundation for AI agents. Here’s why those decisions mattered more than anyone predicted.
-
2026-02-05
OpenClaw and the Shift from Chat to Delegation
I installed OpenClaw expecting a weekend experiment. What I found was something closer to delegated execution, and a clear lesson in why trust boundaries matter when your AI agent has shell access.
-
2026-01-15
From AI Pilot to Operating Model
88% of AI pilots never reach production. The ones that do share a pattern: they rebuild workflows from scratch instead of automating the old ones, and they treat the cultural shift as the actual work.
-
2025-09-19
The Wrong First Question in Azure AI
The Responses API going GA in Azure AI Foundry isn’t primarily a model story. It’s a runtime architecture decision that most teams make too late.
-
2025-07-12
An 18-Month ATO Doesn't Make Software Secure. Here's What Does.
The traditional DoD ATO process was sold as a security control. It was actually a snapshot, useful once and immediately aging. SWIFT changes the fundamental model, and the industry response tells you everything about how ready the field is.
-
2025-02-04
Stop Wiring Apps Together. Build a Domain Reasoning Layer Instead.
Most teams using AI function calling are optimizing at the wrong level. The wins aren’t app-by-app; they’re domain-wide. Here’s what that actually means, why legacy systems make it hard, and how the infrastructure is catching up.
-
2024-04-20
The Compliance Gate Is Open. Now Build Actual Architecture.
Azure OpenAI’s FedRAMP High P-ATO cleared the biggest compliance bottleneck federal AI teams faced. What comes after that clearance is harder, and most teams aren’t building it right.