{
  "id": 17753381,
  "title": "AI Can Write Code. Engineers Still Have to Make It Work.",
  "url": "https://urgent.news/2026/11/02/ai-can-write-code-engineers-still-have-to-make-it-work",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-02T05:50:59.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/jyotshana_0905/ai-can-write-code-engineers-still-have-to-make-it-work-3c61"
  },
  "original_language": "en",
  "account": "Marc Benioff and Mark Zuckerberg have spoken boldly about artificial intelligence's impact on software engineering. However, crafting code is merely one facet of the job, and the tools that generate code still require human oversight, evaluation, and enhancement. Imagine an AI coding assistant crafting a patch that appears correct— it compiles, the tests pass. Yet, it might misunderstand a requirement, dismantle an older feature, or mishandle an uncommon scenario. Someone must still recognize this. This is my apprehension when I hear predictions that AI will supplant software engineers. Generating code is not synonymous with delivering dependable software. Engineers must comprehend what people require, fit modifications into intricate systems, test potential failure points, and assume responsibility for the outcome. AI can aid with segments of this work, but it does not extinguish the rest. In late 2024, Salesforce CEO Marc Benioff hinted that the company might refrain from adding software engineers in 2025, citing productivity gains from AI tools. In January 2025, Meta CEO Mark Zuckerberg projected that AI could function as \"a sort of mid-level engineer\" capable of writing code. These were significant assertions, yet they were forecasts concerning productivity and capability, not evidence that enterprises no longer necessitate engineers. These statements alone do not demonstrate that AI can autonomously grasp a product, sustain a large system, and make sound decisions when issues arise. A 2025 study by METR provides one valuable test of the productivity claims. Sixteen seasoned developers completed 246 tasks in open-source projects they were familiar with. On average, when they could employ the AI tools tested, the developers took 19% longer. Oddly, after the study, they estimated that AI had made them swifter. This outcome demands context; it involved a limited study of experienced developers working in familiar, complex codebases, utilizing a specific set of tools in early 2025. It does not affirm that AI perpetually decelerates engineers. Instead, it illustrates that code generation and real-world productivity are distinct concepts, and perceived speed can diverge from quantified completion time. METR's paper elucidates the study and its limitations. Google's DORA research offers a related challenge at the team level. Its 2024 report indicated that heightened AI adoption correlated with reduced software delivery throughput and stability in the examined organizations. This is an association, not proof that AI incited the decline. The report proposes one plausible reason: AI may expedite code generation, while substantial changes can take longer to review and render systems less stable. Rapid output still necessitates rigorous testing and review. This is why engineers retain their indispensability even when AI composes more of the code. Individuals must elucidate the goal, furnish context the model may lack, appraise whether the result aligns with the need, and convey mistakes and user feedback back into the system. This feedback could emanate from a failed test, a bug report, a customer unable to complete a task, or an engineer spotting a security issue during review. The same applies to the AI tools themselves. They hinge on software, data pipelines, computing infrastructure, integrations, monitoring, and safeguards. Engineers construct and sustain those systems. Human feedback aids teams in discerning where an AI feature proves useful, where it falters, and what necessitates alteration. AI will transform engineering work. It might automate certain routine tasks, enable teams to move faster in specific situations, and influence how companies recruit. However, the fundamental question is not merely whether AI can compose code. It is whether it can reliably assist in delivering software that satisfies people's needs and who verifies that it does. For now, the evidence supports a more measured conclusion than \"engineers are being replaced.\" AI can generate code. Engineers ensure that code finds its place within a product, operates within the surrounding system, and serves the individuals who utilize it.",
  "summary": "Marc Benioff and Mark Zuckerberg have made bold predictions about AI and software engineering. But writing code is only one part of the job and the tools that generate it still need people to guide, review and improve their work. Picture an AI coding assistant producing a patch that looks right. It compiles. The tests pass. But it misunderstands a requirement, breaks an older feature, or…",
  "key_points": [
    "AI coding assistants create functional code but require human verification.",
    "Engineers must understand user needs, fit changes into complex systems, and test for failures."
  ],
  "editors_take": "Automating code generation shifts the engineering burden from writing syntax to the more complex tasks of system integration, rigorous testing, and maintaining the stability of large-scale software products.",
  "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."
}