When you’ve ever used a 3D printer, you might recall the wondrous feeling if you first printed something you may have never sculpted or built yourself. Download a model file, load some plastic filament, push a button, and almost like magic, a three-dimensional object appears. However the result isn’t polished and prepared for mass production, and making a novel shape requires more skills than simply pushing a button. Interestingly, today’s AI coding agents feel much the identical way.
Since November, I even have used Claude Code and Claude Opus 4.5 through a private Claude Max account to extensively experiment with AI-assisted software development (I even have also used OpenAI’s Codex in the same way, though not as continuously). Fifty projects later, I’ll be frank: I even have not had this much fun with a pc since I learned BASIC on my Apple II Plus once I was 9 years old. This opinion comes not as an endorsement but as personal experience: I voluntarily undertook this project, and I paid out of pocket for each OpenAI and Anthropic’s premium AI plans.
Throughout my life, I even have dabbled in programming as a utilitarian coder, writing small tools or scripts when needed. In my web development profession, I wrote some small tools from scratch, but I primarily modified other people’s code for my needs. Since 1990, I’ve programmed in BASIC, C, Visual Basic, PHP, ASP, Perl, Python, Ruby, MUSHcode, and a few others. I’m not an authority in any of those languages—I learned barely enough to get the job done. I even have developed my very own hobby games over time using BASIC, Torque Game Engine, and Godot, so I even have some idea of what makes a great architecture for a modular program that could be expanded over time.
Claude Code, Codex, and Google’s Gemini CLI, can seemingly perform software miracles on a small scale. They’ll spit out flashy prototypes of easy applications, user interfaces, and even games, but only so long as they borrow patterns from their training data. Very like a 3D printer, doing production-level work takes much more effort. Creating durable production code, managing a fancy project, or crafting something truly novel still requires experience, patience, and skill beyond what today’s AI agents can provide on their very own.
