The Myths of AI: Why You Shouldn’t Fear the Future of Coding

In recent days, I found myself entangled in an unexpected experiment with two AI systems: ChatGPT and CLAUDE. Initially, my goal was straightforward—I aimed to create some graphics for my Substack newsletter. While PowerPoint seemed like a reliable choice, I thought I’d take a bold leap into the world of AI coding for a more streamlined process.

What I anticipated as a time-efficient project quickly spiraled into something reminiscent of a chaotic game of Whack-a-Mole. Each time I attempted to resolve one issue, a multitude of new challenges seemed to emerge, creating a cascade of complications. The experience could best be summed up with a phrase I became quite fond of during this adventure: “CLUSTER***K.” After several frustrating days, I ultimately gave in to the simplicity of PowerPoint and completed my graphics in a little over three hours.

This experience has led me to reflect on the current capabilities of AI in coding. At present, these systems resemble novice programmers in their early training stages, struggling to piece together even the most basic functionalities. And while there is no doubt they will improve over time—thanks to methods like Reinforcement Learning from AI-generated Feedback—today’s AI still has a long way to go.

Thus, the prevailing fear of an AI takeover may often be exaggerated or even manipulated by those with ulterior motives. Instead of worrying about the potential “AI Apocalypse,” I suggest taking a moment to relax. Pour yourself a glass of your favorite drink—be it Scotch or perhaps a calming herbal tea—and focus on more productive pursuits. After all, no technology is without its limitations, and for now, coding remains very much a human endeavor.