It’s strange to think about that just half a century ago, people were using coding languages like Basic and Fortran to create just the most rudimentary shapes and programs on monochrome screens.
Now we live in the future, and the digital possibilities seem nearly endless. We’ve even moved from the era of deterministic programming into machine learning and far beyond that, to where we’re now considering the advent of artificial general intelligence.
Here’s a good example of the whiplash times that we find ourselves in – a recent episode of one of my favorite podcasts, AI Daily Brief, where Nathaniel Whittemore talks about Anthropic head Dario Amadei and his predictions for the software engineering world.
Specifically, Amodei made a comment at a recent interview with the Council on Foreign Relations, suggesting that AI will soon write 90% of all code.
“It was (Amodei’s) comments about the pace of worker replacement in the tech sector that have gone viral,” Whittemore noted, explaining how Amodei posited “human-level systems” being synthesized by models, and attributing this statement to Amodei:
“I’ve never been more confident than ever before that we’re close to powerful AI systems.”
You could, of course, attribute that comment to quite a number of people.
And when it comes to coding, it seems evident that things are changing, quickly, in a big way.
What is Vibe Coding?
Throughout the podcast, Whittemore kept referring to the practice of “vibe coding,” which is a phrase for a new kind of software development that takes advantage of precisely these kinds of automations.
If you know anything about Claude Code or Cursor, you know that AI is now able to create code in pretty sophisticated ways. Vibe coding is the process of working as the human in the loop at an abstract level – giving AI the instructions, and letting it do the work.
Here’s some of this concept in the words of OpenAI cofounder Andrej Karpathy, whose X post on this has, to some extent, gone viral in the geekosphere:
“I just see stuff, say stuff, run stuff and copy paste stuff, and it mostly works,” Karpathy wrote, suggesting that coders can now “fully give in to the vibes, and forget that the code even exists.” “I ‘accept all’ – I barely touch the keyboard – It’s possible because the LLMs are getting so good.”
In going over this phenomenon, Whittemore credits Riley Brown and other influencers as embracing vibe coding and it’s what it empowers them to do in the middle of what he calls a profound paradigm shift. To balance this, he also mentions pushback from some, for example, at IBM, and the assertion by Mark Cuban that “AI is the tool” and still requires human guidance.
However, Whittemore points out that Sundar Pichai said, in the company’s third-quarter earnings call, that Google’s AI-written portion of the codebase is now up to about 25%.
Enterprise Code and the New Era
Here’s the thing, though – as Whittemore also points out, enterprise code bases are quite large.
They’re too big for programmers to keep in their heads.
People will have to ask questions about these huge codebases, which Whittemore called a “labyrinth.”
“The size, maturity and turnover of modern companies, as well as use of libraries and packages, means huge pieces are and always will be unknown to you,” he says. “You still need to be productive in this labyrinth. Three, AI coding will get more powerful, and with it, the ability for an AI to understand this labyrinth will grow. You simply need to ask, for some reason, ask questions about the code as you go. Trying to keep this much info memorized in your brain will not work.”
Overall design decisions, he suggests, may still be made by humans, but in the end, jobs will change.
What it Means
It seems evident that we’re close to outsourcing most human work to AI.
Taking all of this into account, it’s almost impossible to think that we’re just going to be doing the same things several years from now. The job world is going to look radically different. That’s pretty frightening for any of us who need job security, but the best advice that we get is to stay ahead of what’s happening with new technologies, and that’s where these predictions become really useful.
