核心张力在于：AI，尤其是类代理（agent-like）的工具，究竟是扩张多数使用者的能动性（agency），还是把它收窄成「software-shaped」的抱负。文章勾勒出一套 3-case 的「brutal calculus」：(1) 有 software-shaped 的目标再加上写程式背景，表示你的工作有风险；(2) 没有 software-shaped 的目标也没有写程式背景，会被贴上「Luddite」标签；(3) 有 software-shaped 的目标但没有写程式背景，暂时算是赢家。它主张，这种框架即便让科技产业兴奋，却可能疏离一般人。
要走向主流，AI agents 必须维持一种可运作的「social license」，也就是大众的容忍度，让产业能在监管机关收紧规则之前先行扩张；文章警告，即使只是对缺点的过度感知，也可能危及这种许可，而它认为这些缺点目前正在广泛引发共鸣。推进动能的一个关键限制是文化层面：最喧闹的布道者常与许多人不喜欢的矽谷原型形象绑在一起，因此采用与否不只取决于能力或「killer apps」，还取决于这项技术是否让人感到普遍被赋权，而非菁英化或带有强迫性。
A sharp selloff in software stocks this week, driven by fears that AI could replace broad swathes of existing business models, is framed as an early signal of larger disruptions where the boundary between “productivity boost” and “obsolescence” keeps moving. The piece notes the irony that “learn to code” was long treated as a safe hedge, yet even people who followed that advice can be made obsolete again as AI accelerates. Against that backdrop, AI enthusiasts portray the moment as ideal for rapid building and automation, using new jargon and hustle-laced evangelism.
The central tension is whether AI, especially agent-like tools, expands most users’ agency or instead narrows it to “software-shaped” ambitions. The article sketches a 3-case “brutal calculus”: (1) software-shaped goals plus a coding background means your job is at risk, (2) no software-shaped goals plus no coding background makes you a “Luddite,” and (3) software-shaped goals plus no coding background makes you a winner for now. It argues that this framing can alienate average people, even if it excites the tech industry.
To go mainstream, AI agents must preserve a “social license” to operate, described as the public tolerance that allows industries to scale before regulators tighten rules; the article warns that even an outsized perception of downsides can threaten that license, and it suggests those downsides are currently resonating widely. A key constraint on momentum is cultural: the loudest evangelists are associated with a Silicon Valley archetype many people dislike, so adoption depends not just on capabilities or “killer apps,” but on whether the technology feels broadly empowering rather than elitist or coercive.