This one is conducting exams, and the other has a plan in mind. I’m currently coding on four different projects at once, despite the fact that I’m really just burning tokens,” Yegge says, referring to the expense of creating large language models ( LLM) to generate large chunks of text.
Learning to code has long been seen as the solution to a attractive, safe job in technology. Advancements in coding standards from companies like Google, Anthropic, and OpenAI are now a major threat to destroy that idea. Companies are rifling with information about downsizing or even completely erasing their designer teams, according to X and Bluesky.
When ChatGPT debuted in late 2022, AI designs were worthy of autocompleting small regions of code—a good, if modest step forth that served to speed up applications development. Engineers and non-engineers alike began using the tools to create complete apps and websites as models advanced and developed “agentic” skills that enabled them to use software programs, manipulate files, and access online services. In February, Andréj Karpathy, a well-known AI researcher, coined the term “vibe coding” to describe the steps involved in creating software by giving an AI model a text prompt.
The rapid progress has led to speculation—and even panic—among developers, who fear that most development work could soon be automated away, in what would amount to a job apocalypse for engineers.
At a Council on Foreign Relations event in March, Dario Amodei, CEO of Anthropic, said,” We are not far from a world where AI is writing 90 % of the code. I think we’ll be there in three to six months.” ” And then in 12 months, we might be in a world where AI is essentially writing the entire code,” he continued.
But many experts warn that even the best models have a way to go before they can reliably automate a lot of coding work. While advances in AI may lead to the development of a system that can code just as well as humans, relying too heavily on it could lead to a glut of clumsy and hackable code as well as a lack of programmers with the knowledge and skills required to create good software.
It’s possible that software development work will be automated, according to David Autor, an economist at MIT who studies how AI is affecting employment. Similar to how transcription and translation jobs are quickly being replaced by AI, Autor claims. He notes, however, that advanced software engineering is much more complex and will be harder to automate than routine coding.
Autor goes on to say that the situation may be complicated by the “elasticity” of the demand for software engineering, or how much the market could support additional engineering positions.
No improvement in speed or cost reduction would cause a “mad rush” for the proctologist’s office, according to Autor, “if demand for software was like demand for colonoscopies.” ” But if demand for software is like demand for taxi services, then we may see an Uber effect on coding: more people writing more code at lower prices, and lower wages”.
By the end of the year, Yegge predicts that this is how programming will be conducted. ” And if you’re not doing it, you’re just walking in a race”.
The Vibe-Cooking Gap
Coding message boards are full of examples of mobile apps, commercial websites, and even multiplayer games that have apparently been vibe-coded into being today. Experienced coders, like Yegge, can give AI tools instructions and then watch AI bring complex ideas to life.
Cursor and Windsurf are two of the two AI-coding startups that have sparked a wave of interest in the concept. ( OpenA I is frequently reported to be in talks to buy Windsurf. )
At the same time, the obvious limitations of generative AI, including the way models confabulate and become confused, has led many seasoned programmers to see AI-assisted coding—and especially gung-ho, no-hands vibe coding—as a potentially dangerous new fad.
The notion that AI will replace human coders is overstated, according to Martin Casado, a computer scientist and general partner at Andreessen Horowitz who sits on the board of Cursor. AI excels at creating stunning things, but he excels at putting forth specific errors.
Still, Casado has been stunned by the pace of recent progress. He claims,” I had no idea it would get this good this quickly.” Since assembly was replaced by higher-level languages, this is the most dramatic change in computer science.
Ken Thompson, vice president of engineering at Anaconda, a company that provides open source code for software development, says AI adoption tends to follow a generational divide, with younger developers diving in and older ones showing more caution. Despite all the hype, he claims that despite their unpredictable output, AI tools still do not trust them. He claims that AI’s “uncertainty” makes it too dangerous and risky.
Both Casado and Thompson see the vibe-coding shift as less about replacement than abstraction, mimicking the way that new languages like Python build on top of lower-level languages like C, making it easier and faster to write code. The popularity of programming has typically increased as a result of new languages and the number of speakers. AI could, in turn, increase the number of individuals skilled in writing code that is functional.
Bad Vibes
The vibe-coding boom suggests, paradoxically, that having a solid understanding of coding is still crucial. People who work in the field frequently report having problems, including introducing unforeseen security flaws, developing features that only simulate real functionality, unintentionally charging high prices with AI tools, and end up with broken code without any idea how to fix it.
” AI]tools ] will do everything for you—including fuck up”, Yegge says. Like toddlers, you need to watch them closely.
Daniel Jackson, a computer scientist at MIT, predicts that” AI will undoubtedly change the way code is produced.” He is currently looking into how to incorporate AI into large-scale software development. However, it wouldn’t surprise me if the hype would pass if we were deceived.
Jackson cautions that AI models are fundamentally different from the compilers that turn code written in a high-level language into a lower-level language that is more efficient for machines to use, because they don’t always follow instructions. An AI model may take an instruction and perform it better than the developer, but it might perform much worse.
Jackson adds that when someone is creating complex software, the vibe of coding gets worse. ” There are almost no applications in which’ mostly works’ is good enough”, he says. You care about a piece of software working correctly as soon as you care about it.
Changes to one line of code can cause issues elsewhere in the system because many software projects are complex. Experienced programmers are good at understanding the bigger picture, Jackson says, but “large language models can’t reason their way around those kinds of dependencies”.
To accommodate AI blind spots, Jackson thinks software development might evolve with more modular codebases and fewer dependencies. He anticipates that AI will eventually replace some developers and prompt many more to reevaluate their methods and put a greater emphasis on project design.
Too much reliance on AI may be” a bit of an impending disaster”, Jackson adds, because” not only will we have masses of broken code, full of security vulnerabilities, but we’ll have a new generation of programmers incapable of dealing with those vulnerabilities”.
How to Code
Even businesses that have already incorporated coding tools into their software development process claim that the technology is still far too unreliable for use in future.
Christine Yen, CEO at Honeycomb, a company that provides technology for monitoring the performance of large software systems, says that projects that are simple or formulaic, like building component libraries, are more amenable to using AI. Even so, she claims that the developers at her business who employ AI in their work have only increased their productivity by about 50 %.
Yen goes on to say that AI just frankly isn’t good enough yet to be additive for anything that requires good judgment, where performance is important, or where the resulting code touches sensitive systems or data.
” The hard part about building software systems isn’t just writing a lot of code”, she says. Engineers are still required, at least for the time being, to possess that curation, judgment, guidance, and direction.
Others believe that there will be a shift in the workforce. ” We are not seeing less demand for developers”, says Liad Elidan, CEO of Milestone, a company that helps firms measure the impact of generative AI projects. We are seeing fewer developers with average or low performance, according to the report.
Naveen Rao, VP of AI at Databricks, a company that assists large businesses in building their own AI systems, says,” If I’m building a product, I could have needed 50 engineers, but now maybe I only need 20 or 30.” ” That is absolutely real”.
Rao contends that learning to code should continue to be a valuable skill for some time. It’s like saying,” Don’t teach your child to learn math,” he says. Understanding how to get the most out of computers is likely to remain extremely valuable, he adds.
The veteran programmers Yegge and Kim think the majority of developers will be able to adapt to the incoming wave. The pair give new recommendations for software development in their book, Vibe coding, which include modular code bases, constant testing, and plenty of experimentation. Yegge says that using AI to write software is evolving into its own—slightly risky—art form. It’s about how to accomplish this without erasing your hard drive and putting money in your bank account, he says.
