Vibe coding is the future
Who wrote your code, anyway?
Earlier this year, Andrej Karpathy coined the term “vibe coding,” a code‑generation style where developers prompt large language models (LLMs) conversationally - sometimes even by voice - to produce working software with minimal manual coding. He describes it as “giving in to the vibes, embracing exponentials, and forgetting that the code even exists.”
And he’s not alone:
Of a recent Y Combinator batch, 1/4 of the founders said that more than 95% of their code base was AI generated (by LLMs). A year ago, these founders would have built their products from scratch…now they are doing it using GenAI.This is the future of product, according to several senior partners at Y Combinator, who recently sat down to discuss vibe coding. They concluded that:
Vibe coding is the ‘dominant’ way to code.
It's not a fad; vibe coding is not going away.
If you aren't doing it - you might just get left behind.
They did caveat that while vibe coding is the future of building product, founders using vibe coding must develop strong debugging and systems‑thinking skills, as scaling these prototypes is still challenging. And security is still a big concern.
Check out the video here. A few implications of ‘vibe coding’ that we captured:
We're all product people now.
People with technical minds who are systems thinkers can become programmers much quicker than they could have a decade ago (the same kinds of people were joining boot camps back then).
Vibe coders still need to have taste and a knowledge of what is good vs. bad code.
At scale, your code + infrastructure matter much more than during product creation. Can vibe coders go into the details to fix issues? Or do they need to hire someone else?
Expertise is still a differentiator. To become the "best" at coding, vibe coders still need to engage in deliberate practice to get into the details (classically trained computer scientists have the edge here).
Adventure on.