On a screen in Kirkland, Washington, Steve Yegge, an old hand in software engineering, watches as AI churns out code like a machine on steroids. He’s juggling multiple projects, but really, he’s just burning money on AI-generated text. Here’s what this really means: the tech industry’s shiny new toy is showing off, but it’s not the first time we’ve seen this circus.
Learning to code used to be the golden ticket to a solid career in tech. Now, with AI models from the likes of OpenAI, Anthropic, and Google, that ticket might be losing its luster. Social media is buzzing with talk of companies cutting down their developer teams or even considering getting rid of them completely.
When ChatGPT showed up in late 2022, it was a modest step forward, helping to speed up software development. Fast forward, and AI’s now dabbling in building entire apps and websites. Andrej Karpathy calls it “vibe coding”—a fancy way of saying you prompt AI with text and hope for the best.
Developers are getting jittery, fearing an AI takeover that could spell disaster for their jobs. Dario Amodei, CEO of Anthropic, predicts AI will be writing most of the code soon. But let’s not get carried away. The reality check here is that while AI is impressive, it’s got a long way to go before it can reliably replace human coders. Until then, we might just end up with a mess of buggy, hackable code and a shortage of developers who actually know what they’re doing.
David Autor from MIT warns that while some tech jobs might get automated, advanced software engineering is a tougher nut to crack. The demand for software could swell like the taxi industry did with Uber, but it won’t necessarily mean better pay for everyone.
Yegge, once skeptical about AI’s role in coding, now believes it’s the future. He’s even writing a book about it. But here’s the rub: for every coder jumping on the AI bandwagon, there’s another who’s wary of its unpredictable output. AI can do some dazzling things, but it’s not great at specifics. It’s like handing a toddler a hammer and hoping they’ll build a cabinet.
Some see AI as just another step in the evolution of coding, like moving from assembly language to Python. AI could make programming more accessible, but it’s not going to replace the need for solid coding skills anytime soon.
The vibe-coding craze highlights a key point: knowing how to code is still crucial. AI might produce some impressive results, but it’s also prone to producing disasters. It’s not just about writing lots of code; it’s about understanding the bigger picture and ensuring everything works seamlessly.
Christine Yen from Honeycomb points out that AI is far from reliable for critical projects. It might boost productivity for simple tasks, but when it comes to sensitive systems, AI just isn’t up to scratch yet. The real challenge in software development isn’t cranking out code—it’s about judgment, guidance, and direction.
We’re not seeing less demand for developers, just less demand for mediocre ones. Companies might need fewer coders, but they’ll still need the sharp ones. Learning to code remains a valuable skill, much like learning math—it’s about getting the most out of computers.
Yegge and Kim, seasoned coders, suggest that developers can adapt to this AI wave. They recommend strategies like modular code bases and constant testing. Using AI to write software is becoming its own art form—one that requires careful handling to avoid costly mistakes.
