← Home

Writing

Notes on production AI systems: MCP, RAG, agentic workflows, and full-stack architecture.

Building an MCP Server for a Production SaaS: Lessons from Exposing an Experimentation Platform to LLMs

20/07/2026

Tool-first design, auth for LLM clients, idempotent writes, and the schema/prompt failures that show up once MCP hits production.

MCPLLMAgentic WorkflowsTypeScriptAPI DesignAI SafetyRead →

Agentic Write Actions: Letting an LLM Touch Your Production Database Without Losing Sleep

13/07/2026

The safety ladder from read-only to scoped autonomous writes, RBAC, audit trails, confirmation tokens, and a near-miss that forced the design.

Agentic WorkflowsAI SafetyLLMRBACSystem ArchitectureRead →

RAG Beyond the Demo: Shipping Retrieval Over Messy Internal Documents

06/07/2026

Chunking for datasheets vs prose, hybrid search for product codes, golden-question eval before you touch a vector store, and extension latency budgets.

RAGLLMEvaluationInformation RetrievalTypeScriptRead →

Cloudflare Workers AI vs Hosted LLM APIs: When Edge Inference Actually Wins

29/06/2026

Cost, latency, and quality from a real ingestion pipeline, plus a confidence-gated routing pattern between edge models and frontier APIs.

CloudflareWorkers AILLMEdge ComputingCost OptimizationRead →

Building Scalable Error Handling Systems: How We Reduced Support Tickets by 47%

01/02/2025

A deep dive into architecting production-grade error boundaries that capture 99.8% of failures, refreshed for streaming LLM UIs.

ReactNode.jsError HandlingSystem ArchitectureWeb DevelopmentSoftware EngineeringStreamingRead →

System Architecture Patterns for React/Node Applications: Building for Scale from Day One

09/02/2025

How proper architectural decisions early on can 3x developer velocity and prevent technical debt.

ReactNode.jsSystem ArchitectureSoftware EngineeringWeb DevelopmentTypeScriptBest PracticesRead →