Writing / Field notes
The notebook
Thinking out loud about AI, cloud infrastructure, product craft, and the systems in between.
Featured · Nov 1, 2026
Always-On AI Agents: From the Agent Loop to a Digital Colleague
The agent loop is powerful but ephemeral — close your laptop and it's gone. Always-on agents keep that loop and wrap it in durable infrastructure the model never sees: a server-side runtime, a sandboxed cloud computer, persistent memory, and event triggers. Here's how I've come to think about the architecture that turns a terminal tool into a colleague you text.
Read article
Azure as a Platform Product: A Technical PM's Blueprint
A platform PM's job on Azure comes down to three words: reliable, secure, cost-efficient. Azure hands you three levers — the ARM control plane that every request passes through, idempotent Infrastructure as Code, and a four-level scope hierarchy where governance is inherited. Here's how I've learned to read that surface as a product manager, not a certification candidate.
14 min read
Spec-Driven Development: The PRD Grows Up
When an AI agent is the one reading your requirements, the prose PRD stops being a plan and starts being a liability — spec-driven development is what replaces it.
14 min read
Omarchy: What an Opinionated Linux Teaches You About Product Taste
Omarchy is DHH's 'omakase' Linux — Arch plus Hyprland plus a thick layer of curated config that collapses a 10–20 hour setup into a five-question install. I dig into what's real, which marketing claims (runs on any old laptop, installs in under five minutes) need hedging, and what its opinionated defaults teach about product taste.
14 min read
The 'System One Model' Is Not a Model — It's a Router
The most consequential decision in a modern AI product isn't which model you pick — it's whether, for each request, the system should think fast or think slow. The real artifact worth building is the router that decides per request.
14 min read
Your Platform Team Has Customers: Treating Internal Infrastructure as a Product
An internal developer platform lives or dies on product discipline, not technical merit. The developers who use it are customers — and they will route around anything that is harder than the workaround.
17 min read
Building Nibblify: What I Learned Turning Long Tutorials Into Bite-Sized Daily Episodes
A while back I shared Nibblify here when it was barely an idea — a sketch of a thing I wished existed. Since then I actually built it, shipped a desktop version, threw away one of my core assumptions,…
11 min read
Building a Personal Fitness Coach: Wiring My Fitbit Air Into a Local Coaching App
TL;DR: I got tired of my Fitbit Air data just sitting in an app, so I wired it through Google Health into a local Flask app that reads my steps, sleep, resting HR, and HRV, runs them through a rules-b…
12 min read
Skills Used in AI Agents
Everyone is building AI agents right now. Startups, enterprises, solo developers — if you work in software, you've either shipped one or you're being asked to. But after spending the last year and a h…
11 min read
N8n: The Workflow Automation Platform That's Quietly Becoming Enterprise AI's Backbone
I've spent the last few months digging into the workflow automation space, trying to answer a question that keeps coming up in conversations with engineering leaders: "Should we be using an AI agent f…
13 min read
The AI-Native SDLC: How AI Is Rewriting Every Phase of Software Development
In 2024, I watched a junior engineer on my team ship a feature in two hours that would have taken me a full day in 2022. They weren't faster typists. They were working with an AI coding agent that dra…
13 min read
The $4.4M Question: Why Technical PMs Can't Afford to Skip AI Governance
Last quarter, I watched an AI feature I'd been championing get pulled from our roadmap three weeks before launch. The model worked. The accuracy was solid. The integration was clean. What killed it? W…
13 min read
The Web Developer's Guide to MCP: Connecting LLMs to Browser Tools & APIs
By Mukund Murali · September 2026 --- If you've built web applications for any length of time, you know the pain of integrations. Every third-party API has its own auth flow, its own SDK, its own qu…
12 min read
Agent Harness: The Missing Layer Between Your AI Agent and Production
Hero image — engineering control room with monitoring screens(https://images.unsplash.com/photo-1558494949-ef010cbdcc31?w=1200) Photo by Taylor Vick(https://unsplash.com/@tvick) on Unsplash(https://un…
12 min read
The AGI Era Compute Buildout — Why Infrastructure Is Becoming the Real Bottleneck for Frontier AI Models
Data center server racks — the physical backbone of AI(https://images.unsplash.com/photo-1558494949-ef010cbdcc31?w=1200&q=80) Photo by Taylor Vick on Unsplash(https://unsplash.com/photos/FO7JIlwjOtU) …
11 min read
MCP vs RAG vs Fine-Tuning: How to Actually Choose
A few months ago I watched a team spend three weeks fine-tuning a model so their assistant could answer questions about their product catalog. The day they shipped it, the catalog changed. The model w…
8 min read
Leo: My Personal AI Assistant and Second Brain
I don't manage Leo. Leo manages things for me. --- There's a version of AI that feels like magic, and a version that feels like homework. The homework version is familiar: open a new chat, re-expla…
10 min read
Context Engineering — The Skill That Separates Good AI Users from Great Ones
The skill that separates people who get great results from AI — from those who don't. --- There's a common misconception about why some people get brilliant results from AI tools like Claude or Chat…
5 min read
Interfacing AI: Deciding Between MCP and CLI in Agentic Workflows
As we transition from simple "Chatbots" to autonomous "Agents," the most critical design decision isn't which LLM to use—it's how the agent interacts with the world. For decades, the industry default…
4 min read
Building a Second Brain with Claude + Obsidian: The Ultimate Guide to Compounding Knowledge
AI Second Brain Hero(https://images.unsplash.com/photo-1719550371336-7bb64b5cacfa?fm=jpg&q=80&w=2000&auto=format&fit=crop) Imagine a world where you never "forget" a great podcast, a crucial insight …
4 min read
How to Build a Private, Agentic Workflow with Claude Code and Ollama
In the hyper-accelerated world of AI-assisted engineering, developers face a critical dilemma: intelligence vs. privacy. How do you leverage the elite reasoning of flagship models like Claude without …
7 min read
🧠 Build a No-Code OCR App Using OCI Generative AI and Streamlit
(https://cdn.hashnode.com/res/hashnode/image/upload/v1747195276283/b39d59c0-4a4b-4b41-b1d5-33ec53f65fce.png align="center") 👋 Introduction If you're a developer, cloud architect, or AI enthusiast wh…
3 min read
Unlocking AI Potential: The Model Context Protocol (MCP) Revolution
Model context protocol (MCP) It is a protocol that allows LLMs to access custom tools and services. If your LLM wants to talk to an agent for a specific task then with MCP it is easy to integr…
2 min read
How to Create a Smart Research Assistant Using OCI AI Agents
(https://cdn.hashnode.com/res/hashnode/image/upload/v1741091443260/3e9dfd3b-b2bc-4080-87d3-d5336c46c53e.png align="center") In today's rapidly evolving technological landscape, artificial intelligenc…
8 min read
Exploring Your Cloud with Steampipe and Powerpipe: An Overview with OCI Focus
(https://cdn.hashnode.com/res/hashnode/image/upload/v1731208583927/a95dc693-a9c9-4d39-81fd-20e817690b19.png align="center") This blog post explores Steampipe and Powerpipe, powerful tools for gainin…
5 min read
Google's NotebookLM: A Second Brain for the AI Age
What is NotebookLM? NotebookLM is an AI-powered research tool developed by Google, designed to function as a personalized knowledge base powered by AI. Features and Capabilities: Source Grounded AI…
7 min read
Inbound NAT with Palo Alto in OCI
This topic is with reference in accessing private resources like Load balancer through the untrust Vnic of NVA like Palo Alto. This is a general scenario when we have a firewall like Palo Alto inside …
2 min read
Prompt Engineering 101 - Getting Started
We already know what ChatGPT or Google's Bard is and how it helps us. In a nutshell, we give an input to these AI models and it gives us an corresponding output. This output can be relatively correct …
3 min read