{
  "id": 10932553,
  "title": "I Use AI to Build Software. I Still Don't Trust the Code.",
  "url": "https://urgent.news/2026/09/30/i-use-ai-to-build-software-i-still-dont-trust-the-code",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-30T11:45:33.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/unnita1235/i-use-ai-to-build-software-i-still-dont-trust-the-code-cci"
  },
  "original_language": "en",
  "account": "The author uses AI to create software but admits they do not fully trust the generated code. While AI accelerates the process of turning ideas into working prototypes, identifying whether the resulting code is actually good proves challenging. AI can produce convincing code quickly, even when the underlying logic may be faulty. Developers can inadvertently trust code simply because it works in a particular context, overlooking potential issues such as unnecessary dependencies, duplicated logic, incorrect assumptions, weak error handling, inconsistent patterns, or security problems. The author now adopts a more critical approach, questioning the changes made by AI, the assumptions it made, and its understanding of the developer's intentions. If they cannot explain the generated code, they refrain from blindly accepting it. AI is beneficial for tasks like exploring implementation approaches, generating boilerplate, explaining unfamiliar code, creating initial feature versions, refactoring, writing tests, finding bugs, and generating documentation. However, for significant architectural decisions, the author insists on understanding the consequences before relying on AI. Building AI applications introduces additional uncertainties, as problems can arise at various stages, including the model, retrieval system, tools, database or APIs, retrieved information, and the final response. Therefore, AI system development demands more than just crafting an effective prompt. The author emphasizes that AI is merely one component of the application, and the surrounding system is equally crucial. The overall takeaway is that AI streamlines coding but doesn't eliminate the need for understanding the problem. In fact, it introduces new responsibilities, as developers may overlook implementation details when AI generates them. The author's current practice involves using AI extensively but remaining vigilant not to confuse generated code with understood code, maintaining a workflow that ensures they retain a firm grasp on the underlying logic.",
  "summary": "I use AI to build software. I also don't trust it. That might sound strange considering most of my recent projects involve AI, LLMs, agents, RAG systems, and AI-assisted development. But after building several projects this way, I have learned something simple: Getting AI to generate code is easy. Knowing whether that code is actually good is the difficult part. AI can make you feel much more…",
  "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."
}