The internet had been ablaze with demos of what GPT-3 can do. GPT-3 by OpenAI is a natural language machine learning system trained from 45TB of text data, which allows it to generate sorts of content, all based on just a few input words.
In one particular demo, a GPT-3 integrated into Figma as a plugin can take a URL and a description as inputs, and mock up a website immediately. In the demo, the developer just typed in: “a website like stripe.com that is about a chat app”, and out came a website mockup on Figma looking like Stripe.com, but with a smartphone image showing a chat app, with relevant accompanying text.
In another demo, I saw GPT-3 spitting out React code just based on vague descriptions of what the website that the developer wanted. The amazing thing is that it can interpret a relatively vague description and output something that more or less matches what you are asking.
Just words to website. 🤯
The possibilities are mind-blowing! Imagine, just typing out a few words, and voila, your website is up. A few more sentences, and boom! Your SaaS product is live. Even nocode tools can’t match up to this level of speed to launch. Of course, because it’s trained using books, articles, and other texts that humans wrote, it contains all our human biases too. But while that gets sorted out, I can’t help but feel that the reasons to learn to code to make your own products is getting more distant by the day.
Sure, you still need programmers to make tools like GPT-3, but I’m not interested in making these world-changing things. My ambitions are smaller and simpler, and if GPT-3 can make them, so be it. I’m not beholden to coding as a craftsman who enjoys his craft would. My objective is only to make stuff that I’m interested in.
So... to learn to code, or not to code, that is the question.
And maybe in the future where there's a GPT-10, that becomes a non-question for most people...
Follow my daily writings on Lifelog, where I write about learning to code, goals, productivity, indie hacking and tech for good.