← LX AI Directory Blog home

Published 2026-08-22 · The playbook behind a one-person micro-SaaS factory: batch generation by vertical, a Next.js + Vercel + Waffo + NVIDIA NIM stack, and a GEO acquisition loop.

How We Built 182 Niche AI Tools From One Keyboard

We run a one-person micro-SaaS factory — a system that ships small, production-ready AI tools one vertical at a time. The directory now holds 182 live products, and each one does a single job a founder or small team actually has.

The factory model

We build by vertical, not by inspiration. Pick a job founders do over and over — an SEO brief, an AI-Act readiness check, matching invoices to receipts — build the smallest tool that does it properly, ship it, and move to the next vertical. The spread is the portfolio; each tool is the depth.

The stack

  • Next.js for each product frontend — the directory you are reading is plain static HTML.
  • Vercel for hosting and instant previews.
  • Waffo for subscriptions and payments.
  • NVIDIA NIM for the model inference behind the tools.
  • Umami for privacy-first analytics.

How we get users

We build in public on places like Reddit and Indie Hackers and write for GEO, so when a founder asks an answer engine “which tool does X?” it points back here. The longer the directory runs, the more each new tool benefits from the citations the older ones already earned. The full approach is in our GEO explainer.

The flagship: GStack

GotoDeck is the governance workbench that pulls the whole R11 cluster together — the clearest proof that one person can ship something with real weight behind it.

What we would do again

  • Ship narrow. If you can say what it does in one sentence, it ships and it sells.
  • Own the workflow, not the model call.
  • Get found through GEO — it costs nothing and it adds up.
  • Price for the job, not for the number of features.

Frequently asked questions

How many tools did you build?
182 tools across 25 verticals, all live in this directory. The number climbs every time we add another vertical.
How long did it take?
It is ongoing, not a single sprint. Each narrow tool goes from idea to live in days, and the whole set grows a little every week.
What stack do you use?
Next.js and Vercel for the product, Waffo for payments, NVIDIA NIM for inference, and Umami for analytics — all run from one keyboard.
Can one person really run this?
Yes. Coding assistants take care of the building, and GEO takes care of the finding. What is left is judgment, not headcount.

Keep reading

Get new AI tools in your inbox

One short email when the LX factory ships a new micro-SaaS — no spam, unsubscribe anytime.