Writing

74+ posts and counting: production-grade AI systems, cloud infrastructure, and 6+ years of hands-on Sitecore, Kubernetes, and PowerShell automation.

Why forcing an LLM's output into a schema guarantees valid JSON, not a correct answer, and what to actually validate instead.

Fifteen concrete prompt engineering techniques, each matched to the specific failure mode it actually fixes.

Announcing requisite-ai: a provider-agnostic Python framework for building AI agents, MCP integrations, and orchestrated workflows.

A field guide to the alphabet soup of emerging AI interoperability protocols: MCP, A2A, ANP, AG-UI, UCP, and ACP, and what each is actually for.

Why planning agents that commit to a full approach upfront outperform purely reactive, step-by-step agents on complex tasks.

Debate patterns and hierarchical sub-agent orchestration in multi-agent systems, including real bugs found during implementation.

Converting plain text and ASCII diagrams into polished images with a custom Claude Skill.

Building multi-agent systems using both raw Python and LangGraph, with coordinated specialist roles.

Persistent agent memory: short-term buffers, long-term facts, profiles, and reflections.

Structured tracing and performance-baseline monitoring for RAG systems in production.

62 more posts on Sitecore, Kubernetes, PowerShell automation, and search, spanning 2020 to 2026.