The Lathe and the Large Language Model: Crafting Code in the Age of AI

Since April 2025, I have been working on a new chapter of my artistic and technical journey. My source directory now holds over 715 HTML files across 20 different projects. For those who have used the utilities on my site—like the Vessel Curve Capture or the Spinning Top Designer—I want to pull back the curtain on how these tools were built.

A Career in Data, a Future in AI

I spent 30 years protecting data for large private and government organizations and managing Fortune 100 databases. I earned my Master’s degree in Software Development in 1991, back when the world was transitioning between what we called “generation languages.” Throughout my career, I’ve written in everything from 2GLs like Assembler to 3GLs like C, C++, and Pascal, and 4GLs like SQL and PL/SQL.

To me, working with Generative AI (like Gemini Pro) isn’t a shortcut—it’s the next evolution. If 4GLs were task-specific languages for databases, then AI feels like a 5GL or 6GL. It allows me to describe the “what” and let the machine handle the “how” of the exact syntax.

The “Daft Subordinate” Workflow

I often describe working with an AI as having a “really fast but slightly daft subordinate staff”. It doesn’t always understand me perfectly, and it never complains, but it requires constant oversight. I have found and fixed numerous syntax errors in the code it generates, particularly because the AI must adhere to strict browser security rules that limit how it can access a computer.

My development process is a blend of old-school discipline and modern speed:

  • Version Control: Rather than using GitHub, I manage versions manually using folders on my local machine.
  • Hosting: I use Netlify to push updates, which automatically archives old versions, allowing me to roll back if an “improvement” goes sideways.
  • Safety by Design: Interestingly, AI often “reinvents the wheel” for each project rather than pulling from standard public libraries. This can actually be a security benefit, as it avoids the shared bugs and exploits often found in common third-party code.

Safety and User Vigilance

I take all reasonable measures to ensure my utilities are safe. My IS background means I’m naturally inclined toward data protection, but let’s be clear: no software is “virus-proof.”

Disclaimer: While I review the code, I provide these tools without any implied warranty. Users should always remain watchful and exercise the same caution they would with any third-party web application.

The Ethics of the Craft

A user recently asked if using AI feels “unethical” or like “cheating.” To me, it’s like using a CNC machine to carve a bowl. There is still human intent, setup, and a finger on the “cancel” button at all times. Whether I’m using a gouge on a piece of Baltic Birch or a prompt in a terminal, the goal is the same: to turn an idea into something of value that others can see and use.

Scroll to Top