Blog
Deep-dives into AI agent architectures, RAG optimization, and production AI systems
I stopped typing: the hands-on guide to voice-first work with Wispr Flow, with a one-week list and a Skills kit
5,410 dictations, 1,616 of them into a terminal. Three real examples, ten minutes of Wispr Flow settings (MCP included), a one-week task list, and four open-source Skills you can install today.
I made my portfolio readable by AI agents. An agent-readiness scanner kept me honest.
How I took yonyon.ai from a 36/100 agent-readiness score to 73/100 (grade B, rank #89 of ~11,900) by adding llms.txt, a hand-rolled MCP server, NLWeb /ask, and fixing three embarrassing implementation bugs.
I audited my own mailing stack and found 17 problems
AI agents, one instruction, no sugarcoating. Three of 17 findings were bad: no unsubscribe path, every reply bounced, and no Israeli spam-law layer. The audit-fix-verify loop shipped and verified in prod in under a day.
Building AI Agent Architectures That Scale
A deep dive into multi-agent system patterns, from single-agent scripts to tiered architectures with supervisor agents, tool routing, and memory management.
MCP Servers: The Missing Piece in AI Tooling
How Model Context Protocol servers bridge the gap between AI models and real-world tools. Practical patterns for building MCP servers with TypeScript.
From RAG to Production: Lessons from Shipping AI Systems
Real-world lessons from building and operating production RAG pipelines, chunking strategies, embedding models, retrieval quality, and the infrastructure that makes it all work.