Vibe Coding Origins
Vibe Coding Origins
Adapted from "Vibe Coding in Product Teams: Reconfiguring AI-Assisted Workflows, Prototyping, and Collaboration" by Jie Li, Youyang Hou, Laura Lin, Ruihao Zhu, Hancheng Cao, and Abdallah El Ali (CHIWORK '26). Chapter 1 — "Forget That the Code Even Exists" In early 2025, Andrej Karpathy posted a line that would ripple across the tech world: there is a new kind of coding, he said, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. He called it "vibe coding. " Almost overnight the phrase escaped Twitter/X and lodged itself in how designers, engineers, and founders talk about building software. The idea is deceptively simple. Instead of writing software line by line, you describe what you want in plain language, and an AI system — ChatGPT, Claude Code, Cursor, Lovable, Bolt, and others — translates that intent into working code, interfaces, and interactive prototypes. Earlier AI helpers like autocomplete only sped up small tasks. These new tools behave more like continuous partners, collapsing the once-separate stages of ideation, prototyping, and implementation into a single flowing conversation. But here is the problem the researchers noticed: the hype had raced far ahead of the evidence. Everyone was doing vibe coding, yet almost no one had rigorously studied what it actually does to the way real product teams work, collaborate, and divide responsibility. Does it genuinely make teams faster, or does it quietly introduce new fragilities? Does it empower newcomers, or set traps for them? So Jie Li and her colleagues set out to answer two guiding questions — the intellectual spine of the entire study: RQ1: How do product team members integrate vibe and AI-assisted coding into their design workflows, from ideation through prototyping to implementation? RQ2: What opportunities and risks do product teams envision for human–human and human–AI collaboration in the age of vibe coding? Notice they did not treat vibe coding as pure magic where humans vanish. They adopted a broader, more realistic definition: workflows that are conversational and intent-driven, but that still demand human debugging, verification, and iterative refinement. This distinguishes their work from the fantasy of "pure vibe coding" that spits out disposable throwaway prototypes with zero human oversight. Their core thesis, which the rest of the study defends, is that vibe coding does not merely speed up existing work — it reconfigures it. It reshuffles who participates, what gets validated first, where trust lives, and who gets to claim ownership. That reconfiguration is where all the interesting tension hides. But how do you actually study something as slippery as a "vibe"? In the next chapter, we step inside the method — and meet the twenty-two professionals whose daily struggles became the evidence.