Techwright

The Manifesto

The word developer comes from the French développer — to unfold, to reveal what is already there. For most of software's history, that was literal: you typed what you meant, character by character, and the machine executed exactly what you produced.

That era is ending.

I coined Techwright because no existing term captured what I actually do. I am not a “prompt engineer” — that term implies someone optimizing a text box. I am not a “citizen developer” — that implies someone without real technical depth. I am not “using AI tools” in the way an accountant uses spreadsheet templates.

What a Techwright Is Not

It is not vibe coding. Vibe coding — generating something that looks right and shipping it — is the cheap, fast, often catastrophic version of this. A vibe coder and a Techwright both use AI to write code. The difference is everything that happens before and after.

Before: A Techwright has a complete architectural picture before the first line is generated. They know the data model, the security boundary, the failure modes, the scale requirements. They are not discovering the architecture through iteration — they are executing a plan that exists fully formed before the AI writes a word.

After: A Techwright reads what the AI produced. Really reads it. Not to check syntax — to check judgment. Does this handle the edge case at the authentication boundary? Is this the right abstraction, or does it create a hidden coupling that surfaces six months from now? Would this hold under an adversarial penetration test? A vibe coder ships the output. A Techwright ships after the output has been judged.

Why This Is the Future

The coding job is not disappearing. The coding job is transforming — from manual syntax production to architectural direction and judgment. The engineers who will thrive are not the fastest typists or the best at memorizing APIs. They are the people who can hold a complex system in their head, communicate it precisely to an AI, and evaluate the result against production-grade standards.

In 30 years across the US Navy, Dell, Visa, APCT, and Shield AI, I accumulated something that has no line item on a résumé: architectural judgment. The knowledge of which database schema will fail at scale. The instinct for which security model will collapse under real adversarial pressure. The experience to know that the clever microservices design will become the maintenance disaster in 18 months. This judgment is not replaceable by AI. It is the thing that directs AI.

The Obligation

Techwriting carries specific ethical weight — especially in defense and security contexts. When I directed the development of security systems that real organizations depend on, every AI-generated output was my responsibility. The machine does not bear accountability. The Techwright does.

This is why “ethics in the loop” is not a principle I added for optics — it is the mechanism that prevents Techwriting from becoming a vector for careless, unreviewed software in high-stakes environments. In defense AI, security infrastructure, and compliance systems, the cost of a missed edge case is not a bug report. It is a breach, a failed audit, or a system that lets an adversary through. The Techwright is the human in the loop. That position is not a burden — it is the definition of the role.