Writing
Articles & posts
Notes from twenty years of building enterprise infrastructure: AI, network automation, multicloud, and what actually works at scale.
Why I built netlld: network design docs that don't go stale
A Low-Level Design takes days to write and is out of date the moment the network changes. So I stopped writing them by hand.
Read article →Devin, Claude Code, Cursor or Codex? Product differences and use-cases
They all turn a request into working code, but they're built for different jobs. A simple rule for picking an AI editor, terminal agent, cloud agent, app builder or framework.
Read on LinkedIn ↗AI Quantization
How storing a model's weights in 8, 4 or 2 bits cuts cost and memory, what accuracy you give up, how it differs from distillation, and which level to choose.
Read on LinkedIn ↗Agent Harness
The orchestration layer that wraps an LLM with tools, memory, sandboxes and feedback loops. How it differs from an agent SDK, with Claude Code and Codex as examples.
Read on LinkedIn ↗EPYHIA: I let four AI agents spend money. The hard part was permission.
A four-agent AI agency that turns a plain-language brief into a live site, marketing pack and Stripe checkout, with every action going through one gate. $0.76 for the whole business.
Read on LinkedIn ↗Your AI agent has already been compromised. You just don't know it yet.
Introducing Adversarial Agents, my field manual for securing AI agents, built on the 3C Triad: Context, Capability and Channel. An attacker only has to win one.
Read on LinkedIn ↗Software Doesn't Ship Until Someone Shows Up
The demo worked. That was the problem. What a Forward Deployed Engineer actually does, what the role costs, and how to tell if it's you.
Read on LinkedIn ↗Concierge AI: a multilingual voice agent for hotel reservations
Real-time voice AI in 11 languages with barge-in, cited hotel-policy answers and a head-to-head benchmark. Building the AI was the small part.
Read on LinkedIn ↗The AI second brain: a subnet manager for my own knowledge
How I turned a junk drawer of PDFs and bookmarks into a self-linking LLM Wiki with Claude and Obsidian, then made it agentic. With a 5-minute walkthrough.
Read on LinkedIn ↗The AI Second Brain: LLM Wiki (video)
A 5-minute walkthrough of the LLM Wiki pattern with Claude and Obsidian: from a junk drawer of PDFs to a living, self-linking knowledge base.
Watch on YouTube ↗DebateForge: The Token Economics of Engineered Reasoning
What one round of AI reasoning actually costs: a 5-layer Claude pipeline, 14x token growth, and $0.0432 per debate.
Watch on YouTube ↗My AI productivity stack: Obsidian, Notion, Claude Code and OpenClaw
The first time four tools have felt like one system: think in Obsidian, coordinate in Notion, build with Claude Code, scale with OpenClaw.
Read on LinkedIn ↗AI Productivity Tools: Obsidian, Notion, Claude Code & OpenClaw
Four tools as one connected system, with best practices and a 4-week plan to set them up.
Watch on YouTube ↗4 Certifications That Get You Hired in AI & Cloud in 2026
Built with Zaid for students, career switchers and non-tech beginners. A clear, jargon-free path into AI and cloud.
Watch on YouTube ↗No posts on this topic yet.