---
id: "concept-vibe-coding"
type: "concept"
source_timestamps: ["§ 1. Zero-latency iteration."]
tags: ["software-engineering", "prompt-engineering"]
related: ["concept-zero-latency-iteration", "entity-org-atomic", "entity-org-anterior"]
definition: "AI-assisted software development where users generate and modify functional code using natural language prompts, enabling non-engineers to build software."
enrichment_verdict: "Partially supported — non-engineers can prototype and modify software with LLMs, but the claim that this completely bypasses engineering is an overstatement for production-grade systems."
sources: ["futures"]
sourceVaultSlug: "hbr-seg-futures"
originDay: 2
articleStem: "hbr-new-24-agentic-ai-supercharges-startups"
sourceUrl: "https://hbr.org/2026/07/how-agentic-ai-supercharges-startups-and-threatens-incumbents"
sourceTitle: "How Agentic AI Supercharges Startups and Threatens Incumbents"
---
# Vibe-Coding

Vibe-coding is AI-assisted software development that uses **natural language prompts** rather than traditional syntax-based programming. It lets founders and non-engineers — such as UX designers — rapidly prototype functional front-end user interfaces and modify software code directly, bypassing traditional engineering bottlenecks and blurring the boundaries between distinct product-development roles.

Concrete deployments in the source: [[entity-org-atomic]] used vibe-coding to prototype a functional UI in *days*; at [[entity-org-anterior]], UX designers change UI code without engineers. It is a key enabler of [[concept-zero-latency-iteration]].

**Enrichment note.** Research and product evidence strongly support natural-language programming and UI-building (MIT Sloan: agents perform tasks like writing contracts or determining prices at much lower marginal cost). However, robust systems still typically require engineering oversight for architecture, security, and maintainability. *Verdict: Partially supported — non-engineers can prototype and modify, but 'completely bypassing' engineers is not best practice for production systems.*