[ilaksh]

My take on this is that we have multiplied the number of potential ways to build a web application continuously for decades. And now there are a combined 10,000 (pick a number) viable (but not trendy) ways to make a web application.

The interesting thing to me is that ultimately businesses don't care how it works. They just want it to work.

Which means that you can pick a small framework or two across the frontend and backend and configure or train an AI system on only a tiny fraction of the sum total of web development knowledge and have an effective automated web development system.

Looking at the trajectory of gpt4 to gpt-4o and Llama2 to llama3, the prevalence of multimodality, improved reasoning ability as models get bigger, strong investment in hardware research, etc.

I've been doing web development in some capacity since the late 90s and focused on leveraging generative AI for the last two years. I don't see how any reasonable person can follow this stuff closely and not anticipate AI systems that can literally do the entire job of a small web development team, within just a few years. It was actually possible to build a version of that two years ago, and some of the latest attempts are very polished, if lacking in some level of functionality. But that is coming.

Every single job that we have today will be automated. I assume they means that people will be left basically herding swarms of AI agents. For a few years. But it won't be very long before you really need an AI to control your agent swarms or even understand what they are doing.