Samuel Beek，来自阿姆斯特丹的创作者在使用 ChatGPT 指引下自行制作电动门开门器后，导致家中所有保险丝全部熔断，这次失败让他转向更有结构的硬体开发模式。对他而言，在硬体领域关键差异是二分的：不是把保险丝烧断，就是出货一个可靠的产品。Schematik 将以终端使用者为导向，将文字提示转化为可实际组装的设计、零件清单与组装指引，正是为了弥补这个差距。

Beek 将这个概念做成了应用程式 Schematik，反复描述它是「Cursor for Hardware」。该服务对实体设备的作用类似于 vibe coding 在软体中的作用：使用者描述目标，系统回传元件、接线与可购买渠道建议。Beek 已开始规划商业化并筹资；该公司最近获得了来自 Lightspeed Venture Partners 的 460 万美元。2026年2月在 X 上发布后，许多制作者进行了测试，其中包括 N8N 品牌负责人、同时也是投资人 Marc Vermeeren，他分别做了 MP3 播放器与名为 Clawy 的 Tamagotchi 风格机器人。同期 Anthropic 在 X 宣布推出小型 Bluetooth API，让开发者建立可与 Claude 互动的硬体；社群反应亦显示，模仿与竞争已在发生。

文章描绘了更大的趋势：AI 公司进入硬体的速度与云端软体巨头与晶片公司相似，而创客文化早已从可穿戴到各类改装设备广泛存在。Beek 认为硬体仍被门槛高的限制所困，核心在于零件复杂度与相容性，尤其是 SKU 层级对位的困难。他将安全定位为核心约束，故将 Schematik 限于低电压构建，3 至 5 伏特即可，足以支援物联网设备与音乐播放器，但不足以支援高风险系统；他仍表示长期目标可能包含人形机器人。iFixit CEO Kyle Wiens 认为大量零件与 BOM 匹配是一种 AI 擅长处理的规模化优化问题。Beek 进一步对比了发展节奏：过去 5 年软体进步极快，而硬体在过去 10 到 20 年几乎没有同等规模的进展。

Samuel Beek, an Amsterdam-based maker, moved from a dangerous first attempt to a structured hardware-coding model after he fused every circuit in his home with a self-built electric door opener guided by ChatGPT. He now says that in hardware, the difference is binary: you either blow fuses or ship a reliable product. The user-facing goal of Schematik is exactly this gap: turn text prompts into physically buildable designs, parts lists, and assembly guidance.

Beek converted that concept into the app Schematik, repeatedly describing it as a “Cursor for Hardware.” The service performs for physical devices what vibe coding has done for software: describe the goal, then receive recommendations for components, wiring, and where to buy them. He has started monetization planning and fundraising; the company recently received US$4.6 million from Lightspeed Venture Partners. After an X post in February, makers tested it broadly, including Marc Vermeeren (N8N branding lead, also an investor), who built an MP3 player and a Tamagotchi-style bot called Clawy. In the same week, Anthropic announced a small Bluetooth API for building Claude-connected hardware, while community sentiment suggests competition and imitation are already unfolding.

The article frames a larger trend: AI makers are entering hardware as aggressively as cloud software leaders and chip firms, while maker culture already spans devices from wearables to niche retrofits. Beek argues hardware remains gatekept by component complexity and compatibility, especially SKU-level mismatch. He positions safety as the core constraint, limiting Schematik to low-voltage builds (3–5 volts), enough for IoT gadgets and music players but not dangerous systems; he says long-term ambitions still include humanoids. Kyle Wiens of iFixit sees scale and BOM matching as an AI-suitable optimization problem. Beek adds a structural contrast: software improved dramatically in the last five years, while hardware has seen little comparable change over ten to twenty years.