The short answer
andrej karpathy named it in february 2025: describe what you want in plain english, let the AI write the code, run it, see what happens, iterate. you're the director. the model is the one typing.
The short version
a founder i talked to last month shipped his first prototype in 72 hours. no technical co-founder. no agency. just him, cursor, and claude. he was proud of it, until the app broke the moment his second user signed up simultaneously.
that story is vibe coding in 2025, exactly as it is.
not a scam. not a revolution. something more specific, and more useful, than either of those takes.
what vibe coding actually is
andrej karpathy named it in february 2025: describe what you want in plain english, let the AI write the code, run it, see what happens, iterate. you're the director. the model is the one typing.
that's it. there's no certification. no specific tool. the definition is that loose.
what matters for founders is what sits underneath the definition. vibe coding flips the assumption that building software requires a person who can read and write code fluently. for most of the last 30 years, that assumption was basically correct. right now, it isn't, not entirely.
i'm not saying engineers are obsolete. i'll come back to that. but the gap between "i have an idea" and "i have a working thing i can show someone" has compressed from months to days for a specific class of product. that compression is real and it changes what's possible for a non-technical founder with limited runway.
the honest version of what it's good at
vibe coding is fast when the problem is contained.
landing pages, internal tools, simple CRUD apps, early prototypes you're showing to 10 potential customers. These are the right contexts. the AI understands the scope. the iteration loop is short. you describe a change, you see it, you move on.
at apexstack, we use AI-assisted development across almost every project we take on. it's not that we hand everything to an LLM and walk away. it's that the right AI tooling, used by someone who understands the architecture, compresses the timeline on well-defined work in a way that would've been impossible three years ago. that's how we get a production-ready MVP out in 4 to 6 weeks instead of 4 to 6 months.
the distinction matters. vibe coding as a practice isn't "let AI build everything." it's "use AI to close the gap between intent and working software, faster than before."
where it breaks. And it does break
i thought vibe coding meant founders could skip the technical layer entirely. i was wrong. it means the technical layer moves, not disappears.
here's where it predictably fails:
Frequently asked questions
- anything that needs to scale
- AI writes code that works. it doesn't always write code that works for 10,000 users simultaneously. the database queries that look fine at 50 rows fall apart at 500,000. the architecture that made sense for a demo becomes a liability in production. this isn't theoretical. It's the exact failure mode i see founders hit when they vibe-coded their MVP and suddenly had traction they weren't prepared for.
- security and auth
- an AI will implement authentication. it will also, without careful review, implement it in ways that leave serious vulnerabilities, exposed API keys, insufficient rate limiting, broken session handling. this isn't a knock on the models. it's a knock on the assumption that "it generated code" means "the code is safe." it often isn't, not without someone who knows what to look for.
- complex integrations
- connecting stripe is fine. connecting stripe + a third-party KYC provider + a custom webhook flow + a real-time notification system, where the state has to be consistent across all four. That's where vibe coding runs out of context window and patience simultaneously. the model loses track of what it built three prompts ago. the codebase becomes something nobody fully understands, including the AI.
- anything you'll need to maintain
- code that was vibe-coded without structure tends to be code nobody wants to touch six months later. no documentation, no consistent patterns, logic scattered across files because that's how the prompts happened to go. if you're building something you plan to grow, the short-term speed of vibe coding can create a technical debt ceiling that kills you later.
Talking about product engineering?
A long-running team that owns discovery, delivery and iteration — for products that keep evolving after launch.