Security
Why Your Internal MCP Server is a Security Liability (And How to Fix It)
MCP is a communication pipe, not a security boundary. Secure internal tools with least privilege and the boring success criterion.
Older articles
Newest first
Technology
Home Lab to Client-Ready: Drawing the Line Between Tinkering and Production Support
The gap between a functional home lab and production-grade support is procedural discipline, not hardware.
Automation
Where Playwright Ends and AI Agents Begin
Deterministic automation is hitting a wall. I’ve spent years debugging selector failures and anti-bot blocks. Here is why hybrid AI agents are the
Automation
Stop Automating Publishing: A Resilient Workflow Guide for WordPress
WordPress automation fails when it ignores human editorial friction. Learn to build resilient workflows that prevent duplicate drafts and ensure quality.
Security
The Operator’s Guide to SMB Security: Why Frameworks Fail and Habits Win
Stop buying tools. Start building habits. A practical guide to SMB security: MFA, patching, verified backups, and access reviews.
AI Tools & Reviews
Feature Flags for AI Automation: Kill Switches, Confidence Thresholds, and Rate Limits
A practical guide to implementing AI feature flags, kill switches, and quality gates for safe, cost-controlled AI rollouts.
Developer
Why We Self-Hosted Langfuse: The Operational Reality for Small Teams
Escaping SaaS lock-in isn't just about cost; it's about control. We detail the real tradeoffs of self-hosting Langfuse and OpenTelemetry, from
AI Tools & Reviews
The Seven-Field Standard for Production AI Audit Trails
Stop treating AI as a black box. Implement the seven-field standard for audit-ready AI agents that capture prompts, tool calls, and human approvals.
AI Tools & Reviews
Why Your RAG Prototype Is Lying to You: The Case for Rigorous Pre-Launch Evaluation
Prototypes lie by being too clean. Learn to stress-test RAG systems for silent failures before customers see them.
Developer
Stop Duct-Taping AI Agents: Cloudflare Workflows and Durable Objects in Practice
I stopped trying to force multi-step agents into raw Workers. Here’s how Cloudflare Workflows and Durable Objects actually handle state, cost, and failure
AI Tools & Reviews
OpenAI Agents SDK Sandbox Patterns: Safer File and Command Automation
OpenAI's new sandbox architecture separates orchestration from execution, enabling safer, stateful AI agents for enterprise automation.
Developer
Why I Stopped Building Custom AI Routing (And What I Use Instead)
Managing AI infrastructure as a small team is a trap. Vercel AI Gateway solves the sprawl, but only if you accept its architectural constraints....