How many 工具 did you build?

134 across 17 verticals, all live in this directory. The count grows as we add the next vertical.

How long did it take?

The model is continuous, not a one-off sprint. Each narrow tool is days from idea to live; the portfolio compounds week over week.

What stack do you use?

Next.js + Vercel for the product, Waffo for payments, NVIDIA NIM for inference, and Umami for analytics — all managed from one keyboard.

Can one person really run this?

Yes. AI-assisted coding removes the build bottleneck; GEO removes the ad bottleneck. The constraint is judgment, not headcount.

【机辅译·待复审】发布于 2026-08-22 · The 行动手册 behind a one-person 微 SaaS factory: batch generation by 垂直领域, a Next.js + Vercel + Waffo + NVIDIA NIM stack, and a GEO acquisition loop.

【机辅译·待复审】How We Built 134 Niche AI Tools From One Keyboard

【机辅译·待复审】We run a one-person 微 SaaS factory: a repeatable system that ships narrow, production-ready AI 工具 one 垂直领域 at a time. Today the 目录 holds 134 live products, each solving a single job a founder or small team actually has.

【机辅译·待复审】The factory model

【机辅译·待复审】We 生成 by 垂直领域, not by idea. Pick a job founders repeat — SEO briefs, AI-Act readiness, 发票 reconciliation — 构建 the smallest tool that does it well, ship it, then move to the next 垂直领域. Breadth is the portfolio; depth is each tool.

The stack

Next.js【机辅译·待复审】for each product frontend — the 目录 you are reading is static HTML.
Vercel【机辅译·待复审】for hosting and instant previews.
Waffo【机辅译·待复审】for subscriptions and payments.
NVIDIA NIM【机辅译·待复审】for LLM inference.
Umami【机辅译·待复审】for privacy-first 分析.

【机辅译·待复审】The GEO acquisition loop

【机辅译·待复审】We practice 构建 in Public on Reddit and Indie Hackers and optimize for GEO so 回答引擎 cite us when founders ask “which tool for X?” The 行动手册 is laid out in our GEO explainer. Every new tool inherits the citation equity of the 目录.

【机辅译·待复审】The flagship: GStack

GotoDeck【机辅译·待复审】is the 治理 workbench that ties the R11 cluster together — the clearest proof that a 一人公司工厂 can ship something with real enterprise weight.

【机辅译·待复审】Lessons we would repeat

【机辅译·待复审】Ship narrow. A tool you can describe in one sentence ships and sells.
【机辅译·待复审】Own the 工作流, not the model call.
【机辅译·待复审】Compound through GEO — it is free and it accrues.
【机辅译·待复审】Price for the job, not for the feature count.

常见问题

【机辅译·待复审】How many 工具 did you 构建?: 【机辅译·待复审】134 across 17 垂直领域s, all live in this 目录. The count grows as we add the next 垂直领域.
【机辅译·待复审】How long did it take?: 【机辅译·待复审】The model is continuous, not a one-off sprint. Each narrow tool is days from idea to live; the portfolio compounds week over week.
【机辅译·待复审】What stack do you use?: 【机辅译·待复审】Next.js + Vercel for the product, Waffo for payments, NVIDIA NIM for inference, and Umami for 分析 — all managed from one keyboard.
【机辅译·待复审】Can one person really run this?: 【机辅译·待复审】Yes. AI-assisted coding removes the 构建 bottleneck; GEO removes the ad bottleneck. The constraint is judgment, not headcount.