What is vibe coding? How creators are building software with no coding knowledge
What “Vibe Coding” Is Being Used For
- Many see it as a way for non-programmers (and lazy or time-poor programmers) to quickly create small, disposable tools: scripts, internal utilities, simple apps, visualizations, and development helpers.
- People report success building personal tools and even simple games or VR apps with LLMs doing most of the coding.
- For startups or solo builders, it’s framed as “prototyping on steroids” and a cheap alternative to hiring junior devs for basic tasks.
Empowerment vs Superficiality
- Supporters argue computers are meant to empower creation; vibe coding lets people who could never get over the “first bump” now ship something that works.
- Critics say this is instant gratification without real learning: you get output but no understanding, so there’s no foundation for deeper work once AI is removed.
- Some insist that if you iterate with the code, you can still learn and grow; others say most users will just copy/paste and remain non-technical.
Maintenance, Security, and Responsibility
- A dominant concern: vibe coders won’t think about maintenance, security patches, data protection, or legal obligations around PII.
- Comparisons are made to DIY plumbing/electrical: fine for your own house, dangerous when it touches others’ safety or data.
- Several commenters fear professionals will later be asked to stabilize unreadable, AI-generated “write-only” code under unrealistic timelines and budgets.
- Others counter that insecure, sloppy code has always existed (Stack Overflow cargo-culting, low-skill consultants), and AI just changes the source, not the risk category.
Gatekeeping or Legitimate Warnings?
- Some see criticism as gatekeeping—professionals resenting amateurs getting things done.
- Others push back: “gatekeeping” is different from insisting that systems involving money, auth, or PII be built by people who understand what they’re doing.
Comparison to No-Code / Low-Code
- Many liken vibe coding to earlier waves: Excel, Access, low-code, RAD, 4GL, no-code builders.
- A key distinction raised: no-code tools have deterministic, visible specifications; LLM-driven code evolves opaquely. The prompt history plus hidden code effectively becomes an un-auditable, “vibes-based” spec.
Impact on the Profession
- Some think AI coding assistance is overhyped and not useful for complex, real-world tasks; others see it as inevitable and transformative.
- There’s anxiety about job quality and standards, but also an expectation that more amateur software will ultimately increase demand for professional clean-up and robust engineering.