The backlash against vibe coding is really a fight over whether AI-assisted software work is still a novelty or already a serious engineering discipline.
Boris Cherny has touched a nerve by saying he is sick of the phrase vibe coding. That might sound like a small complaint about tech slang, but the reaction around Claude Code shows something larger is happening. A term that once made AI prototyping feel playful is now being tested against production systems, investor decks, hiring plans and the credibility of an entire category.
According to Business Insider, Cherny, the creator of Claude Code at Anthropic, has been looking for a better name for AI programming after arguing that vibe coding feels too casual for tools that are now doing serious commercial work. The timing matters. The phrase, popularized after Andrej Karpathy used it in early 2025, spread because it captured the strange new feeling of describing software in natural language and watching an AI agent turn intention into code. It was funny because it was true.
Now the joke is wearing thin. A Reddit thread asking for alternatives drew more than 300 comments in roughly an hour, which is not normal terminology chatter. That kind of velocity usually means people are arguing about status, not vocabulary. Engineers hear vibe coding and worry it makes their work sound like guessing. Founders hear it and see a phrase that sells speed, accessibility and the promise that small teams can build more with fewer people.
For startups, this is not a cosmetic issue. Naming shapes what investors believe a market is becoming. If a founder says their platform enables vibe coding, the pitch can sound fast, approachable and viral. It can also sound unserious, especially to buyers who have to think about security reviews, audit trails, maintainability and compliance. That gap is where many AI developer tool companies now have to operate.
The early market rewarded demos. A nontechnical founder could prompt an app into existence, record the screen, post the result and make the future feel obvious. That was useful. It introduced millions of people to the idea that software creation would no longer be limited to those who could write every line by hand. But demos are not the same as products, and the industry is now moving into the less glamorous phase where the output has to survive other users, changing requirements and real customers.
This is why engineers are pushing back. They are not rejecting AI coding tools. Many of them use Claude Code, Cursor, Codex and similar systems every day. The objection is that vibe coding collapses too many different activities into one phrase. There is a meaningful difference between prompting a throwaway landing page, refactoring a legacy service with tests, reviewing an AI-generated migration, and supervising agents across a large codebase. A single playful label makes all of that sound like the same thing.
That matters in hiring too. Startups are already rewriting expectations for technical roles. The strongest engineers in this environment may not be the ones who type the fastest, but the ones who can decompose problems, inspect generated work, design guardrails and know when the model is confidently wrong. Calling that vibe coding undersells the judgment involved. It also gives weaker candidates a convenient way to describe output without showing engineering discipline.
The market is searching for a grown-up vocabulary
Cherny's own search for a replacement is revealing because the obvious alternatives are awkward. Agentic engineering sounds more serious, but it also sounds like something built for a conference slide. AI-assisted development is accurate, but dull. Prompt engineering already carries baggage and does not capture the broader workflow. Software engineering may still be the right umbrella, but the tools are changing the day-to-day job enough that people clearly want a new phrase.
The best replacement will probably not come from branding alone. It will come from practice. If teams build durable systems with AI agents, the vocabulary will shift toward the workflow that wins. That may mean words around orchestration, supervision, verification or AI-native engineering. The exact phrase matters less than the standard it implies: developers are not just asking a model for code, they are designing a process where machines generate work and humans remain accountable for outcomes.
There is also a commercial reason toolmakers should care. Enterprise buyers do not want to buy vibes. They want faster shipping without a security incident, cleaner maintenance without hidden technical debt, and productivity gains that do not collapse when the first edge case appears. The companies that sell AI developer tools will have to explain how their products fit into disciplined teams, not just how quickly they can spin up a prototype.
Still, the phrase is unlikely to disappear overnight. Vibe coding works because it is memorable, and memorable language has a way of surviving professional embarrassment. The likely outcome is a split. Casual builders and social media will keep using the term for fast, prompt-driven creation, while serious teams adopt more precise language for production workflows. That split may be healthy, because it lets the market separate play from practice.
The bigger signal is that AI coding has moved past the phase where novelty can carry the story. The next winners will be companies that make AI-generated software more reliable, reviewable and secure, not just more spectacular in a demo. Watch the language founders use over the next few months. When the market starts naming the discipline differently, it usually means the discipline itself is changing.
Also read: Monk is using rent as a recruiting weapon for AI talent • Small Companies Are Turning AI Hype Into Market Fuel • Singapore gives students a practical route into building with AI
