We’re One Script Away From 30% Cheaper Groceries, Stop Asking AI, Start Asking Why

我们距离杂货价格降低 30% 只差一个脚本，别再问 AI 了，问问为什么吧

You don’t need a smarter model to make food cheaper. You need to delete a middleman. That’s the whole argument. Everything below is just the receipts. 你不需要更聪明的模型来让食品变便宜，你需要的是剔除中间商。这就是全部论点，下文所列皆为证据。

The tax nobody votes on: Open any grocery or delivery app. The first three results aren’t the closest store, or the cheapest one. They’re the ones who paid to be there. That’s not an accusation, it’s the business model. Delivery platforms commonly take 15–30% commission off every order. Sponsored placement is its own line item on top of that. None of it comes out of the platform’s pocket, it comes out of the price you pay, baked quietly into the menu before you ever see it. 无人投票的“隐形税”：打开任何杂货或外卖应用，前三个搜索结果既不是离你最近的店，也不是最便宜的店，而是那些付费买下排名的商家。这不是指控，这是商业模式。外卖平台通常会从每笔订单中抽取 15%–30% 的佣金，推广位费用还要另算。这些成本并非由平台承担，而是直接转嫁到你支付的价格中，在你看到菜单之前，这些费用就已经悄悄地加进去了。

Multiply that across every grocery store, every restaurant, every small food business that wants to be findable online, and you get a strange situation: the discovery layer, the part that just tells you a store exists and what it sells, has become one of the most expensive parts of the entire food supply chain. Not farming. Not logistics. Being found. Take that tax out, and prices don’t need a subsidy to drop. They just drop, because the markup that was funding it isn’t there to fund anymore. 将此推广到每一家杂货店、每一家餐厅以及每一家希望在网上被发现的小型食品企业，你会发现一个奇怪的现象：发现层（即告知你商店存在及其售卖商品的部分）已成为整个食品供应链中最昂贵的环节之一。不是农业，也不是物流，而是“被发现”。只要去掉这笔税，价格无需补贴就能下降。它们会直接降价，因为支撑这笔费用的溢价已不复存在。

We think that number is north of 30% for a meaningful slice of the market. Call it a working thesis, not a peer-reviewed study, but it’s the kind of thesis you can actually go test, because the fix is small enough to build in a weekend. The fix isn’t AI. It’s a map nobody owns. 我们认为，对于市场中相当大的一部分而言，这个溢价比例超过了 30%。这只是一个工作假设，而非同行评审的研究，但这是一个你可以亲自去验证的假设，因为解决方案简单到只需一个周末就能构建出来。这个方案不是 AI，而是一张无人拥有的地图。

Here’s the uncomfortable part: the technology to remove the discovery tax isn’t new, exotic, or AI-shaped. It’s boring. It’s: A public list of where stores are and what they sell; A way for anyone to read that list without paying for the privilege; A way for store owners to edit their own listing for free. That’s it. That’s the whole disruption. No model, no inference cost, no API bill. A map and a catalog, both open, both free to read and free to write. 令人不安的事实是：消除这种“发现税”的技术并不新颖、不奇特，也不涉及 AI。它很枯燥，仅仅是：一份关于商店位置和售卖商品的公开列表；一种让任何人都能免费读取该列表的方式；一种让店主免费编辑自己列表的方式。仅此而已，这就是全部的颠覆。没有模型，没有推理成本，没有 API 账单。一张地图和一个目录，两者都是开放的，且读写免费。

This is exactly what we open-sourced last week with the Food Freedom Map, a pin-drop map built on Nostr where any store, farmer, or restaurant can list themselves and their catalog, and any client can read that data, forever, without paying anyone for placement. No ad auction. No “boost your listing” button. The map doesn’t rank you by budget, because there’s no budget to rank by. 这正是我们上周通过“食品自由地图”（Food Freedom Map）开源的内容。这是一个基于 Nostr 构建的地图，任何商店、农户或餐厅都可以列出自己及其商品目录，任何客户端都可以永久读取这些数据，无需向任何人支付排名费用。没有广告竞价，没有“提升排名”按钮。地图不会根据预算对你进行排名，因为根本没有所谓的排名预算。

The interesting question isn’t whether this can work, the protocol’s already running. It’s how fast you could migrate an entire region’s worth of existing stores onto it. The migration script, roughly: Say you wanted to bootstrap a city overnight. The shape of it looks like this: 有趣的问题不是这是否可行（协议已经在运行了），而是你以多快的速度将整个地区的现有商店迁移到该平台上。迁移脚本大致如下：假设你想在一夜之间启动一个城市的数据，其流程如下：

Step 1 — Pull the data that already exists. Most of what you need is already public. OpenStreetMap’s Overpass API will hand you structured shop/restaurant/category data for an entire city in one query, no scraping required. Government open-data portals (business registries, food-safety permits, market licenses) fill in the gaps OSM misses. You could also scrape something like Google Maps for the same fields — plenty of projects have — but that’s walking into a Terms of Service fight with a company that actively fingerprints and blocks exactly that kind of traffic. Open data gets you 80% of the way there with none of the legal risk. Start there. 第一步——获取现有数据。你需要的大部分数据已经是公开的。OpenStreetMap 的 Overpass API 可以通过一次查询为你提供整个城市的结构化商店/餐厅/类别数据，无需爬虫。政府开放数据门户（商业登记、食品安全许可、市场执照）可以填补 OSM 的空白。你也可以抓取 Google 地图等平台的数据——很多项目都这么做过——但这会让你陷入与公司的服务条款纠纷中，因为它们会主动识别并封锁此类流量。开放数据可以让你在没有任何法律风险的情况下完成 80% 的工作，从这里开始。

Step 2 — Normalize it into one schema. Name, location, category, that’s the minimum viable pin. This is the boring-but-essential part: every source names things differently, so you write one small mapping layer and you’re done. Genuinely a script, not a system. 第二步——将其标准化为统一模式。名称、位置、类别，这是最小可行性标注。这是枯燥但必要的部分：每个来源的命名方式都不同，所以你只需编写一个小的映射层即可。这确实只是一个脚本，而不是一个复杂的系统。

Step 3 — Drop it on the open map as unclaimed pins. The store exists on the map whether or not the owner has ever heard of your project. That’s the point — discovery doesn’t wait for adoption. 第三步——将其作为未认领的标注点投放到开放地图上。无论店主是否听说过你的项目，商店都会出现在地图上。重点在于——“发现”不应等待“采用”。

Step 4 — Give every pin a free, editable post. This is the part that makes it stick. The owner finds their pin, claims it, and can edit their catalog directly — what they sell, what they’re asking for, what they’re giving away — for free, forever, the same way anyone claims a Google Business Profile today. Except there’s no algorithm deciding who shows up first, and no fee for not disappearing. 第四步——为每个标注点提供一个免费、可编辑的帖子。这是让项目持续下去的关键。店主找到自己的标注点并认领后，就可以直接编辑目录——售卖什么、需求什么、赠送什么——永久免费，就像今天人们认领 Google 企业资料一样。不同之处在于，没有算法决定谁排在前面，也没有为了不被“消失”而支付的费用。

Run that sequence and a city’s worth of stores can go from “invisible unless they pay” to “on the map, editable, free” in roughly the time it takes to write and run an import script. That’s not a five-year roadmap. That’s a weekend project sitting on top of a protocol that already exists. 执行这一流程，一个城市的商店就能在编写并运行导入脚本的时间内，从“不付费就隐形”转变为“在地图上、可编辑、免费”。这不是一个五年的路线图，这是一个建立在现有协议之上的周末项目。

“Use your brain, don’t ask AI” “动动脑子，别问 AI”

The headline isn’t anti-AI for the sake of being contrarian. It’s pointing at something specific: this particular problem doesn’t need a smarter answer, it needs a different incentive structure. No amount of better recommendation algorithms fixes a market where the algorithm’s owner gets paid by the seller, not the buyer. You can’t optimize your way out of a conflict of interest. You have to remove it. That removal is mechanical, not magical. It’s a public dataset, an open protocol, and a free listing. The kind of thing one engineer can ship, not the kind of thing that needs a foundation model behind it. 这个标题并非为了反调而反 AI。它指向了一个具体问题：这个问题不需要更聪明的答案，而需要不同的激励结构。如果算法的所有者从卖家而非买家那里获利，那么再好的推荐算法也无法修复这个市场。你无法通过优化来解决利益冲突，你必须消除它。这种消除是机械性的，而非魔法。它需要的是公共数据集、开放协议和免费列表。这是一个工程师就能交付的东西，而不是需要基础模型支持的东西。

Try breaking this 试着去打破它

The repo’s open, the protocol’s documented, and the gap between “idea” and “your city is on the map” is exactly as small as it sounds. If you want to take a swing at the import-script side of this — OSM importer, open-data normalizer, claim-flow UX — that’s a wide-open, genuinely useful first contribution. 代码库是开放的，协议有文档记录，“想法”与“你的城市出现在地图上”之间的差距确实如听起来那样小。如果你想尝试导入脚本方面的工作——OSM 导入器、开放数据标准化器、认领流程 UX——这是一个非常开放且真正有用的首次贡献机会。

👉 github.com/Djowda/food-freedom-map Fork it. Point it at your city. Tell us what broke. Fork 它，指向你的城市，告诉我们哪里出了问题。