YC startups, Google, Microsoft, and even some Indian firms claim that much of their code is now generated by AI. But the reality leans more toward developers who still prefer writing code manually—a truth often obscured in such announcements.
Interestingly, Visa recently posted a job opening for a staff GenAI engineer, which requires the candidate to know tools like Lovable and Replit. This has been doing the rounds on social media, but the important thing to note is that the first criterion in the job description is still about proficiency in Python and PostgreSQL.
The rise of vibe coding tools like Lovable, Replit, and V0, and the influencer marketing around them has made students of computer science question if there is still a need to learn coding. Everyone from VCs to influencers who made fortunes from code suddenly seem eager to declare programming dead.
But ask the people who actually write code, and the picture couldn’t be more different.
Learn to Code to Vibe Code
“You should still learn to code. And you should start as early as possible,” Santiago Valdaramma, founder of Tideily, wrote recently on X. The backlash to this “coding is dead” narrative is growing louder.
Kevin Naughton Jr put it bluntly: “The people who are telling you not to learn to code are the same people who have made a lot of money from knowing how to code.” This is a growing sentiment amongst developers who are getting into the field but are also listening to people who say that now everyone’s a developer.
The term ‘vibe coding’ has been floating around ever since Andrej Karpathy made it mainstream. Sahil Lavingia recently embraced the idea with his tongue-in-cheek post: “Learn to code to vibe code.” Even Logan Kilpatrick chimed in: “You should probably learn to code.”
And yet, no matter how accessible AI makes it to build apps, most serious engineers aren’t buying the hype that actual coding is obsolete. ThePrimeagen, another influencer on X, said, “No amount of vibe coding is going to prevent you from learning how to program.”
Replit’s founder, Amjad Masad, doesn’t like the term “vibe coding” either. In an interview on Sequoia’s Training Data podcast, he said it “cheapens the possibilities”. His vision runs deeper: a world where AI lets anyone, even kids in rural India, build functional apps by describing them in plain English.
So yes, the barriers to entry have dropped. But Masad acknowledged that when you use Replit, you’re not avoiding code; you’re interacting with it at a different level. The agent unfolds the code in front of you. You’re still responsible for what it does.
Masad had just recently made a bold statement on similar lines. “I no longer think you should learn to code,” he said in a recent post on X, adding to a growing discourse around the role of human programmers in an AI-first future. He urged people instead to focus on creativity and problem-solving skills that he believes will hold value even as AI takes over the mechanics of programming.
The Jobs Are Not Going Away
Sure, AI tools will make coding easier. They already have. But they don’t make it unnecessary. Bob McGrew, VP of AI at OpenAI, also recently weighed in. Replying to a post about building operating systems, McGrew said, “This is why students still need to learn to code. The very best practitioners always understand the entire stack.”
That is what experts have been talking about for a while, but the conversation has slowly transitioned into learning coding just enough to know what to do when a vibe coding tool messes things up. “You just need to be able to tell a computer what to do in a way that it will respond, understand what it’s doing and how to optimise that, and fix it when it’s not working,” Austen Allred, founder of GauntletAI said.
To do any of those things well—understanding, optimising, debugging—you need to have a mental model of how code works. Steven Heidel from OpenAI offered a contrarian view: “Don’t learn to write code, learn to read code.”
Back to Learning to Code
Despite the constant hand-wringing about LLMs making engineers redundant, the market realities say otherwise. Todd McKinnon, CEO of Okta, recently said it’s “laughable” to think there’ll be fewer engineers in the future. Every tech wave—from the PC to mobile—has increased the demand for developers. AI will be no different.
Even Google CEO Sundar Pichai said they’re still hiring “superstar engineers” and that demand has shifted, not vanished. The future engineers aren’t just writing code—they’re orchestrating AI agents, validating output, and designing scalable systems. They’re doing more, not less.
Microsoft CEO Satya Nadella emphasised that demand for developers is growing across healthcare, manufacturing, and education. You may not always be building the next Instagram, but you’ll be building.
Meanwhile, OpenAI’s CEO Sam Altman and Anthropic’s CEO Dario Amodei have expressed caution. Altman believes we’ll need fewer engineers eventually, but not yet. Amodei went a step ahead and predicted that AI would do 90% of coding within six months.
Well, part of it is clickbait. “Don’t learn to code” has a lot more viral potency than “coding will evolve”. Another important part of this is privilege. People who already know how to code can afford to flirt with the idea of not doing it anymore. As @tekbog noted, it’s often the folks who’ve climbed the ladder who are now pulling it up behind them.
And let’s not ignore the business incentive. If you’re building an AI agent that writes code, you want people to believe they don’t need to learn it themselves. That doesn’t make it true.
