I lead engineering teams building software for US hospitals, and I spend a large part of my week working out how AI actually changes the way software gets built. This page brings that writing together in one place: articles from my newsletter The Tech Strategist’s Edge on Substack, my Medium essays, and explainers published here.
The common thread is practical: what works when AI leaves the demo and meets a real enterprise, with real teams, real constraints and real consequences.
AI Harness: making agent work checkable
My open-source Harness Engineering project puts controls around AI coding agents: what they may change, what evidence counts as done, and how work carries into the next session.
The kit includes proof matrices, independent verification, scope checks, session handoffs, and an attempt gate that treats a timeout as an unknown outcome rather than permission to repeat an action.
Explore Harness-Engineering on GitHub · Browse my GitHub work
Runnable demos show weak checks missing defects, a verification gate refusing self-approval, and retry handling after a lost response. The repository also contains a handbook, adoption guide, and a small evaluation with its methods and limitations.
Things I build and teach
Working projects and practical learning resources connect the ideas in these essays to engineering decisions.
Odia Panjika: technology in service of culture
A bilingual calendar platform for Odia traditions, with a developer API and automated daily updates. The project write-up describes a deterministic astronomical core and an optional AI layer for cultural narratives.
Design decision: astronomical fields are computed rather than generated by a language model. AI supplies context without changing the calendar calculations.
Explore the project and architecture · Try the application. The project article notes that free hosting may need time to wake up.
RAG: from explanation to practice
Three free resources, from reading to building: RAG Explained covers retrieval, evaluation and production; the RAG Blueprint Workshop is an interactive 60-minute session; and the RAG Lab lets you build a small RAG app yourself, in your browser or in Google Colab.
Leadership in the age of AI
Leading in the AI Age and The Relevance Flywheel connect technical change with teams, learning, and career judgment.
Start here
AI Concepts
A layered map of modern AI in plain language, from how models learn to agents, retrieval and safety. The best place to begin if the vocabulary is new.
RAG Explained
A complete guide to Retrieval-Augmented Generation, from first principles to agentic GraphRAG, evaluation and production.
The AI Glossary
From Confused to Confident: the AI terms everyone uses and few people explain, defined without the jargon.
Series · 4 parts · August 2026
AI Coding Patterns Cartography
A map of how teams are really using AI to write software in 2026, based on a survey of GitHub projects: the schools of practice that have emerged, why the surrounding harness matters more than the model, the contrarian voices worth hearing, and how an enterprise moves from experimentation to governed delivery.
| Part | Article | Read on |
|---|---|---|
| 1 | Ten Schools of AI Coding: A GitHub Cartography of 2026 The distinct approaches to AI-assisted development, mapped from what practitioners actually publish. | Medium · Substack |
| 2 | The Harness Is the Product: Context Engineering as Enterprise Discipline Why the context, tools and guardrails around a model decide the outcome, and how to treat them as engineering. | Medium · Substack |
| 3 | The Multi-Agent Debate and Other Heresies: Contrarian Voices in AI Coding The dissenting positions on multi-agent systems, AI safety and security that deserve a hearing. | Medium · Substack |
| 4 | From Vibe Coding to Governed Delivery: An Enterprise Adoption Playbook How an organisation turns scattered AI experiments into a governed, repeatable delivery practice. | Substack |
Building reliable AI agents
What it takes for agents to behave consistently in production: harnesses, recovery, type safety, evaluation and design patterns.
- The Reliability Shift: Why TypeSafe AI Changes the Game for Enterprise Automation
The reliability wall every production AI system eventually hits, and why typed, structured outputs change the equation. Medium · Sep 2026 - The AI Agent Harness: Why Your Copilot Fixes Fail (And How to Fix That)
Lessons from building a real application with an AI coding agent, and why fixes kept slipping. Substack · Sep 2026 - The Agent’s Lost Harness: Why Your Instructions Fail in New Sessions
Why an agent forgets what you taught it between sessions, and how to build a recovery layer. Medium · Sep 2026 · also on Substack - The Harness, Not the Model: What Engineering Leaders Still Miss About AI Coding
Why chasing the newest or cheapest model misses where the real gains in AI coding come from. Medium · Sep 2026 - Building Evals for AI Agents Is Hard; Here Is the Practical Playbook
Why most teams skip agent evaluation, why that’s a mistake, and a practical way to start. Medium · Jul 2026 · also on Substack - Agentic Engineering Patterns: From Code as Expensive to Code as Infrastructure
What changes in engineering practice when writing code stops being the scarce resource. Medium · Jul 2026 - Beyond the Prompt: Master the 5 Essential Agentic AI Design Patterns with LangGraph
Five core agent design patterns, explained and built with LangGraph in Python. Medium · Jul 2026
AI adoption, leadership and careers
What AI means for teams, organisations, business models and the people building careers around it.
- Your team already adopted AI. They just didn’t tell you.
Most enterprise teams think they haven’t adopted AI. Their developers already have. Substack · Aug 2026 - SaaS Isn’t Dying. The Seat Is.
Reflections from a panel on Agentic AI vs. SaaS at Agile Network India, and what it means for how software is priced. Substack · Aug 2026 - The Relevance Flywheel: A Practitioner’s Framework for Staying Sharp in the Age of AI
A repeatable way to keep your skills current while the ground keeps moving. This blog · Jul 2026 - The 2026 AI Survival Guide: Which Tool to Choose and How to Master It
A practical user manual for the age of specialised AI agents. Medium · Mar 2026 - The Practical AI Skills Roadmap for 2026
What actually matters, and how a beginner can get started without drowning in hype. Medium · Feb 2026 - Future-Ready Engineer: A Step-by-Step Guide for Young Engineers in India
A roadmap for engineers starting their careers as AI reshapes the profession. This blog · Apr 2026 - Experience Changes How You Prompt
Why years of domain experience show up in the questions you ask an AI. This blog · Mar 2026 - Drowning in Abundance: My Life at the Edge of the AI Age
A personal account of living and working amid an overload of AI tools and possibilities. This blog · Feb 2026
Hands-on and builds
- RAG Lab: Build a RAG App in Your Browser
A free, hands-on lab: build a small RAG app step by step, with chunking, hybrid search, filters, citations, safe refusals and evaluation. Runs in your browser with no sign-in, or as a full notebook in Google Colab. This site · Oct 2026 - RAG Blueprint Workshop
An interactive 60-minute training on RAG, from foundations to agentic GraphRAG and production, with labs, quizzes and facilitator notes. This site · Oct 2026 - Build a Python Coding Agent in 30 Minutes
A step-by-step walkthrough for building your own coding agent. This blog · Feb 2026 - Introducing Odia Panjika (ଓଡ଼ିଆ ପଞ୍ଜିକା)
Bridging Odisha’s cultural rhythm with modern technology. This blog · Aug 2026 - The Chronicles of Chad GPT-Prompt: A Modern Engineering Tale
A lighter take on engineering life alongside AI. This blog · Aug 2025
More writing across Medium and Substack
Harnesses, context, and practical AI skills
- The Rise of Harness Engineering: Why the “Wrapper” Is the New Moat
Why the infrastructure surrounding a model matters. Medium - The most expensive thing you can do with an AI is forget how it remembers it.
A reflection on AI memory and context. Medium - The AI Bottleneck Nobody Is Talking About Isn’t the Model
Lessons from building an astrology application. Medium - The Builder’s Toolkit: AI Skills Every Developer Needs in 2026
Practical foundations for developers. Medium - The Governor’s Framework: LLMOps, Evaluation, and AI Governance in 2026
Operating and evaluating AI systems. Medium - The Leader’s Toolkit: AI Strategy, Team Design, and Transformation in 2026
The organisational decisions behind AI adoption. Medium - The Watch List: SLMs, Reasoning Models, and the Frontiers Worth Your Attention
Which developments merit attention and experimentation. Medium - The Agent That Doesn’t Get Lost
A reading of MicroSkill Architecture for enterprise agents. Substack
Strategy, privacy, and the human side of work
- Your AI Does Not Know What It Wants: Decision Theory, Utility, and the Hidden Architecture of Modern Intelligence
Objectives, uncertainty, and the values behind AI decisions. Substack - The Final Frontier of Privacy
An introduction to confidential computing and data in use. Substack - The Waiting Room
A story about fear, feedback, and the human side of leadership. Substack - The Stones We Carry
A personal reflection on self-doubt; also published on this blog. Substack
Get the next one in your inbox
New technology essays go out first on The Tech Strategist’s Edge. Subscribe there, or follow along on Medium.