{
  "id": 19536861,
  "title": "Stop Vibe Coding: How Real Engineers Use AI Without Creating Slop",
  "url": "https://urgent.news/2026/11/11/stop-vibe-coding-how-real-engineers-use-ai-without-creating-slop",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-11T19:04:25.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/jouwdan/stop-vibe-coding-how-real-engineers-use-ai-without-creating-slop-2em1"
  },
  "original_language": "en",
  "account": "Software discussions are currently fixated on \"vibe coding,\" a process where developers describe an application in plain English and an AI model generates the code, which can then be shipped without manual syntax work. While this approach works for quick scripts, landing pages, and prototypes, it falls short when handling real customer data, concurrent traffic, and long-term changes. When engineers rely solely on AI without a technical plan, several problems tend to recur. For instance, when an edge-case bug appears, engineers may paste the stack trace back into the AI chat window, causing the model to patch the symptom but break adjacent modules and make the underlying state messier. Over time, architectural consistency disappears, leading to a single repository with inconsistent state management conventions and duplicate data-fetching patterns.\n\nOperational safety is often overlooked, with concurrency limits, error boundaries, data integrity checks, and authentication edge cases rarely addressed because the AI prompt only verified whether the happy path renders correctly. The resulting output might look like a functional application, but it behaves like technical debt from day one. The bottleneck in software development is not typing syntax, but rather judgment – determining how data flows, where the system fails, and which abstractions will endure six months of product changes. Treating AI as a magical replacement for engineering judgment results in issues rapidly. The better approach is to treat AI like an entry-level developer on the team. The software engineer's role is to set architecture requirements, schema definitions, and system constraints before anyone writes the implementation.\n\nInstead of unchecked prompting where engineers accept massive multi-file diffs without inspection, they should break large features into discrete, reviewable scopes. This means enforcing test coverage, reviewing changes line by line, and rejecting pull requests that introduce anti-patterns. When bugs arise, engineers must identify the root cause in the system and direct the AI model to the precise logic. Validation is crucial – engineers must check if the browser UI renders locally and use unit tests, integration tests, and edge-case verification. Maintaining a clear codebase with fragment-free patterns, directory structures, schema contracts, and typed interfaces is also essential.\n\nIn the end, the skills that matter in software development have not vanished; they remain as important as ever. Schema design and transactional data integrity, recognizing failure modes (like handling third-party webhooks, database stalls, or state sync issues), and exercising restraint by minimizing dependencies are all still critical. To use AI effectively, engineers must step into the tech lead role: setting technical direction, maintaining code discipline, and reviewing what the tools produce before it reaches production.",
  "summary": "A lot of software discourse right now is fixated on \"vibe coding.\" The pitch is familiar: you describe an application in plain English, an AI model spits out the code, and you ship without touching syntax. That workflow holds up fine for disposable scripts, landing pages, and toy prototypes. Once you build software that handles real customer data, concurrent traffic, and long-term changes, the…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}