Insights
Agentic development, AI system architecture, and the technical conditions a team needs before adopting them.
Model vs. Agentic Harness: The Term That Finally Explains AI Agent Governance
What an agentic harness is, why two products running the same model behave so differently, and why AI agent governance lives in the harness, not the model.
From Decision to Action: Implementing Tool Calling, Idempotency, and Observability in AI Agents
How to connect an AI agent to real systems without arbitrary access: schema-validated tool calling, idempotency keys, state separation, and audit traces with OpenTelemetry.
Autonomy Is Not Authority: The Missing Principle in Most Agentic AI Architectures
An AI agent can reason correctly and still lack the authority to act. Risk-tiered autonomy (L0-L5), human-in-the-loop design with LangGraph, and when a multi-agent architecture isn't worth it.
Agentic development: what the real evidence says and what must change in your team before adopting it
What agentic development actually is, what the METR and DORA studies say about its measurable impact, and what technical conditions a repository needs before an agent adds any value.
Vibe Coding in Production: Code That Compiles Is Not Code That Was Fine
What happens when AI-generated code reaches production without anyone who actually understands it reviewing it: the security, architecture, and governance risks 2026 evidence documents, and the minimum controls a team needs before trusting it.