Thinking out loud

Writing.

Essays on the systems I build, the models I work with, and the craft of making software that serious teams depend on.

17 entries, newest first.

---

• September 25, 2026 · 26 min read

  LifeLedger, Part I: Ask Your Past

  A birthday in Karachi. A wedding guest list. A diary entry from thirty years ago. LifeLedger begins with a simple wish: ask a question of the life you have already documented.

  Go · RAG · Ollama · PostgreSQL

• September 20, 2026 · 20 min read

  RAG, Agentic RAG, and Multi-Agent RAG—Explained Like a Human

  A grounded, human guide to three ways AI systems find, reason over, and act on knowledge—from one careful search to a coordinated team of specialists.

  RAG · Agentic AI · Multi-Agent Systems · AWS Strands

• September 20, 2026 · 16 min read

  Before Capital Moves: Euthyna and Governed Disbursement

  A plain-English guide to governed capital disbursement: how Euthyna turns project evidence, rules, approvals, and exceptions into a defensible answer about whether the next release is eligible to proceed.

  Euthyna · Capital Governance · Project Finance · Assurance

• September 20, 2026 · 22 min read

  Engineering with Ollama: A Six-Week Hands-On Course

  An end-to-end curriculum for moving from local model experiments to a governed AI gateway with streaming, structured output, RAG, tools, routing, observability, security, and failure engineering.

  Ollama · AI Engineering · Go · RAG

• September 20, 2026 · 14 min read

  Ollama in Production: The Architecture Around the Model

  Ollama is an inference runtime, not the whole application. This reference architecture separates the client, AI gateway, retrieval, tools, storage, and specialized models into governable boundaries.

  Ollama · AI Architecture · RAG · Agentic AI

• September 20, 2026 · 9 min read

  Streaming AI Responses: SSE, WebSockets, and the Right Transport

  Streaming is a behavior, not a protocol. Here is how SSE, WebSockets, chunked HTTP, and Ollama’s model stream fit together—and how to choose without overengineering.

  Streaming · SSE · WebSockets · Ollama

• September 20, 2026 · 6 min read

  From Catalog to Control Plane: Building an Ollama Model Registry

  A practical field guide to discovering Ollama models, inspecting what they claim to support, and validating what they can actually do before routing production work to them.

  Ollama · Local AI · Model Registry · LLM Operations

• August 19, 2026 · 9 min read

  Claude Code Sessions vs. Context: Why It Remembers Things You Deleted

  I deleted 111 old Claude Code sessions and freed 660 MB. Then a brand-new session casually mentioned a bug from months ago, and I went digging into how sessions, /resume, memory, and context actually fit together.

  AI · Claude Code · Developer Tools · Agentic AI

• April 8, 2026 · 6 min read

  Stop Prompting. Start Engineering.

  Most developers use AI coding tools the way they use a search engine. Boris Cherny’s framework replaces that guesswork with environment design, context isolation, and automated verification.

  AI · Claude Code · Developer Tools · Agentic AI

• January 23, 2026 · 11 min read

  Training and Inference Generation process of LLM

  An easy, farmer-friendly walk-through of how an LLM is trained (by learning patterns from lots of text) and how it later answers your question (by predicting the next small piece of text, one step at a time).

  AI Basics · Tokenization · LLM · Inference

• January 19, 2026 · 8 min read

  Choosing the Right AI System: Simple Models, Reliable Workflows, and Smart Agents

  Not all AI works the same way. Some answer once and stop. Some follow strict instructions. Others think, adjust, and try again. This guide explains AI system designs in plain language—so you can choose wisely, not blindly.

  AI · LLMs · Agentic AI · System Design

• January 16, 2026 · 30 minutes

  Building a Live Job Search Agent with Strands, Ollama, and MCP

  A step-by-step guide to building a live job-search AI agent using Strands for orchestration, Ollama for local inference, and the Model Context Protocol (MCP) to access real-time data from Dice without hardcoded APIs.

  AI · Ollama · MCP · Python

• October 21, 2025 · 11 min read

  The Magic of Shrinking AI: From Giants to Specialists

  How massive AI models are compressed into fast, efficient specialists—and why both giants and small models still matter.

  AI Basics · Quantization · SLM · LLM

• October 18, 2025 · 8 min read

  Small Language Models (SLMs): Why Smaller is Often Smarter

  In the AI world, we’ve been told that bigger is always better. But for most businesses, a massive model is overkill. Here is why Small Language Models (SLMs) are becoming the secret weapon for real-world applications.

  AI · Machine Learning · SLM · LLM

• October 16, 2025 · 4 min read

  The Genius of Cheat Sheets: Understanding LoRA Fine-Tuning

  How do you teach a giant AI new tricks without spending a fortune? Let's debug the magic of Low-Rank Adaptation (LoRA) using the power of 'cheat sheets'.

  GenAI · Machine Learning · MLOps · Fine-Tuning

• August 10, 2025 · 4 min read

  More Than a Mariner: The 250-Year-Old Standard for Leadership

  Exploring the famous 'Qualities of a Naval Officer' attributed to John Paul Jones, and why its core message on character and critical thinking is more relevant today than ever.

  Leadership · History · Principles · Management

• May 23, 2025 · 5 min read

  The Great Naming Bug: Why Greenland is Icy and Iceland is Green

  Is it a historical marketing trick or a climate anomaly? Let's debug the mystery behind the names of these two North Atlantic neighbors.

  Nature · History · Climate · Trivia