{
  "id": 12544118,
  "title": "The new engineering bottleneck isn’t writing code, it’s trusting it",
  "url": "https://urgent.news/2026/10/07/the-new-engineering-bottleneck-isnt-writing-code-its-trusting-it",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-07T04:00:42.000Z",
  "source": {
    "name": "e27",
    "slug": "e27",
    "url": "https://e27.co/the-new-engineering-bottleneck-isnt-writing-code-its-trusting-it-20261004/"
  },
  "original_language": "en",
  "account": "The engineering bottleneck isn't just about writing code, it's about trusting that code. Despite AI's promise of faster coding, a new study reveals an increase in both faster code delivery and instability. Google's 2025 DORA report, based on responses from over 1,100 engineers, shows this contradictory outcome. While AI speeds up the initial coding process, it also increases software delivery instability, meaning teams are moving faster but breaking more code.\n\nThis \"verification tax\" is the cost of AI-generated code, where the time saved during initial coding is often spent on reviewing and verifying the AI's output. In fact, 30% of developers in the study reported little to no trust in AI-generated code, which is understandable given that reviewers spend more time checking the AI's work.\n\nAdditionally, a study by GitClear found that the share of refactored code has fallen from 25% in 2021 to under 10% by 2024, while the share of copy-pasted code has risen from 8.3% to 12.3%. This shift indicates that more code is being duplicated rather than refined, leading to a harder-to-maintain codebase as it grows faster.\n\nFurthermore, a Stack Overflow survey of over 49,000 developers found that while 80% have adopted AI tools, trust in their accuracy has dropped from 40% to 29%, and overall favourability towards AI has dropped from 72% to 60%. The top frustration cited was dealing with almost-right AI output, which makes debugging more time-consuming.\n\nUltimately, the responsibility for bugs lies with the person who reviewed, approved, and merged the code, not just the AI. This shift in responsibility requires engineering teams to budget review time based on the impact of AI-generated changes, not just the number of lines involved.",
  "summary": "For the past two years, the story engineering teams told about AI was simple: it writes code faster, so teams ship faster. That story is only half true. Code does get written faster. But Google’s 2025 DORA report, based on analysis of over 1,100 open-ended responses from Google software engineers, found that higher AI adoption […] The post The new engineering bottleneck isn’t writing code, it’s…",
  "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."
}