这是一场仍在进行的公开实验：能不能把「做产品」变成流水线，让每个产品只承担一个最小目标——每月赚到 1 美元。以下是截至目前的真实数据与结论，包括失败的部分。 An ongoing public experiment: can product-building become a production line, where each product carries one minimal goal — earn one dollar a month? Below are the real numbers and conclusions so far, including the failures.
我们停止投入的 14 个产品里，最主要的死因不是质量，而是渠道错配：给线下餐饮老板做的菜单工具，无法通过我们拥有的任何渠道触达他们；给招聘方做的 JD 工具，需求频次低到无法形成搜索流量。判断一个小工具值不值得做，「我们够不够得着这批人」这一条的权重，高于产品本身有多好。 Across the 14 products we stopped investing in, the dominant cause of death was not quality but channel mismatch: a menu tool for restaurant owners we had no way to reach, a job-description tool for a need that occurs too rarely to generate search traffic. When judging a small tool, "can we reach these people at all" outweighs how good the product is.
周报生成、邮件润色、致辞代写——这类「提示词包装」产品全部被我们砍掉。活下来的产品都有一个共同点：要么整合了模型之外的能力（例如起名同时实时查询域名是否可注册），要么把一件事拆成用户自己不会拆的多步结构（例如合同审读=红旗分级+条款提取+谈判话术+提问清单）。 Weekly-report writers, email polishers, speech generators — every "prompt wrapper" was cut. The survivors share one trait: they either integrate capability the model does not have (checking domain availability while generating names), or decompose a task into a multi-step structure users would not build themselves (a contract review is red-flag triage plus term extraction plus negotiation scripts plus a question list).
最初所有产品统一收 1 美元，这是省事而非策略。现在按场景分层：正在损失金钱的处境（拒付、押金、理赔、签证）定价更高且转化预期更好；高频重复使用的改为月订阅；用量差异极大的改为按次计点；情绪与礼物场景改为免费加自愿打赏——给悼词标价本身就是一种冒犯。当前 31 个在营产品的模式分布：买断 18、订阅 5、点数 4、打赏 4。 Everything started at one dollar, which was convenience rather than strategy. Now pricing follows the situation: cases where money is actively bleeding (chargebacks, deposits, claims, visas) carry higher prices and better expected conversion; habitual use became monthly subscriptions; wildly uneven usage became credits; emotional and gift moments became free-with-tips — putting a price on a eulogy is itself an offence. Across 31 live products: 18 one-time, 5 subscription, 4 credits, 4 pay-what-you-want.
生产线成熟后，单个产品从选题到上线的成本降到了以小时计。但产能提升并没有带来收入——因为货架摆满了，店门口却没有路。这是本实验目前最重要的教训：在分发能力没有验证之前，增加产品数量不会改变结果。 Once the line matured, a product went from idea to live in hours. That did not produce revenue — the shelves filled up while the street outside stayed empty. This is the experiment's most important lesson so far: before distribution is proven, adding products does not change the outcome.
开工前打分（渠道可达性、需求硬度、是否免疫免费替代、付费者是否等于受益者、定价与场景是否匹配、能否产出天然内容），不过线不开工。上线后只看真实漏斗：浏览→免费使用→点击购买→付费。免费使用满 100 次仍零成交，触发关闭审查。每一次关闭都会记录原因并公开。 Before building: a scorecard (audience reachability, need intensity, immunity to free substitutes, whether the payer is the beneficiary, price-to-situation fit, and whether the product yields natural content). Below the line, we don't build. After launch: only the real funnel — views, free uses, checkout clicks, payments. A hundred free uses with zero sales triggers a shutdown review. Every shutdown is recorded with its reason and published.
在营产品 31 个，已归档 14 个，累计真实收入：仍然是零。分发刚刚开始，社交渠道的第一批内容发出于 2026 年 7 月下旬。下一个检查点是首笔真实付款——在此之前，我们冻结新产品上线，把资源全部投向分发与转化。 31 products live, 14 archived, cumulative real revenue: still zero. Distribution only just began, with the first social content published in late July 2026. The next checkpoint is the first real payment — until then, new launches are frozen and everything goes into distribution and conversion.