Is It Time to Retire ‘Vibe Coding’? The Debate Explained

Investor Christoph Janz says it is time to retire the term 'vibe coding', because building with AI is now simply building. Others argue the label still matters as a warning about unreviewed, insecure code. Here is the background, the arguments and what the debate means for users, developers and companies.

Christopher Ort
By Christopher Ort · 5 min read
Share
Clay-style developer with headphones relaxing in front of a laptop while a retirement sign hangs on a neon vibe coding label
On this page
  1. Quick Take
  2. What happened
  3. Background: where the term came from
  4. How the debate breaks down
  5. Pros and cons
  6. What X is saying
  7. What Grok says
  8. Critical analysis
  9. What's still open
  10. About the analysis
  11. i10x Perspective

Few tech phrases have spread as fast as “vibe coding”. Andrej Karpathy coined it in February 2025, Collins Dictionary named it Word of the Year for 2025, and by late 2026 it describes much of how software gets built. Now Point Nine investor Christoph Janz has proposed retiring the term altogether, and X is arguing about whether that is overdue or premature.

i10X Quick Take

Quick Take

  1. The spark: Christoph Janz wrote that it is “time to retire the term ‘vibe coding’”, since saying “I built XYZ” is now understood to mean AI wrote the code.
  2. The pushback: Critics say the label still signals something important: software shipped without understanding or review.
  3. Security angle: Posts in the debate point to scans of AI-built apps with exposed data, and Adobe’s Scott Belsky predicts a swing back to secure SaaS.
  4. Why it matters: The word choice reflects a real question: is AI-assisted building now just software development, or a distinct, riskier practice?

What happened

On November 3, Janz posted that he wants to stop saying “I had Claude/Codex build XYZ”, because “when you say ‘I built XYZ’ it’s understood that you didn’t write the code yourself.” He thanked Karpathy for a “great term for a short era.” The post drew more than 1,700 likes and many replies.

The same day, Scott Belsky, former Adobe chief product officer and investor, posted a “narrative violation”: that preferences will swing back from “DIY vibe-coded software to super secure industry-leading SaaS solutions”, because “while it’s super easy to build anything, endlessly protecting and securing it is super hard.”

Background: where the term came from

Karpathy introduced “vibe coding” in a February 2025 post describing a style of programming where you “fully give in to the vibes” and let an AI write the code. In a retrospective one year later, he called it “a shower of thoughts throwaway tweet.” Collins Dictionary chose it as its 2025 Word of the Year, defining it as using AI prompted by natural language to write code. Since then, tools like Claude Code, Codex, Cursor, Lovable and Bolt have made the practice mainstream, from hobby projects to company prototypes.

How the debate breaks down

PositionArgumentVoices on X
Retire the termAI building is now the default; saying ‘I built it’ is enoughChristoph Janz
Keep it as a warningIt signals unreviewed code and security risk@TrioTop13, @Dino_Codes
Expect a backlashSecuring DIY software is hard; teams return to SaaSScott Belsky

Pros and cons

For users

  • Pro: Non-programmers can build useful personal tools and small apps in hours.
  • Con: Apps built without review may leak data; users of such apps often cannot tell how they were made.

For developers

  • Pro: Huge productivity gains for prototyping and routine code; the stigma of “AI-written” is fading.
  • Con: Dropping the label can blur accountability. Someone still has to own architecture, security and maintenance.

For companies

  • Pro: Faster internal tools and prototypes at lower cost.
  • Con: Shadow IT built by non-engineers creates security and maintenance debt, which is Belsky’s argument for a return to established SaaS.

What X is saying

Security-focused accounts went further. @Dino_Codes wrote that vibe coding “has a security problem, and it’s structural,” citing a scan in which “71% of the Supabase apps in this scan returned data with no login.” i10x could not independently verify that scan.

What Grok says

Grok’s replies on X take a pragmatic line. Asked whether vibe coding can replace software engineering careers, Grok said it “speeds prototypes and can yield freelance or product wins, but production systems still need engineers who own architecture, reliability” and related concerns. That is close to the middle of the debate: the practice is here to stay, but responsibility for the result does not disappear.

Critical analysis

Janz is right that the novelty is gone: for many builders, AI writing the code is the default. But “vibe coding” never only meant “AI wrote it”. In Karpathy’s original sense it meant not reading the code, accepting changes on vibes. That second meaning is exactly what security critics worry about, and it has not become less relevant. A better outcome than retiring the term may be splitting it: “AI-assisted engineering” for reviewed, owned code, and “vibe coding” for throwaway prototypes. Belsky’s prediction is plausible for regulated or security-sensitive software, less so for personal and internal tools.

What’s still open

  • Will companies set policies distinguishing reviewed AI code from unreviewed builds?
  • Will AI coding tools ship stronger default security, such as database access rules?
  • Does the term fade naturally, or keep living as a warning label?

About the analysis

Researched and written with AI assistance under the editorial responsibility of the i10X News team. The title image is an AI-generated illustration. Read our AI Use Policy. Spotted an error? Request a correction.

This independent, research-based i10x analysis is based on posts on X from November 3–4, 2026, Karpathy’s retrospective, and Collins Dictionary coverage. Claims from individual posts are attributed and were not independently verified unless stated.

i10x Perspective

The term matters less than the habit behind it. As AI writes more code, the differentiator is no longer who can build, but who can explain, secure and maintain what was built. Whatever we call it, the useful test is the one posted on X: can you say where your app stores user data and who can see it?

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

CodingNews

Editor's picks

  1. Weight-Space Programming Lets GPT-6 Astra Edit Qwen Weights
  2. Anthropic IPO Filing Includes 5% Equity for Public Benefit Fund
  3. Local LLM Inference Bottlenecks: VRAM and KV Cache Limits
  4. Self-Improving AI Agents Shift to Recursive Scaffolding Updates
  5. Anthropic’s IPO Filing Exposes $895B Compute Commitments to 2045