Why we're building this in public — and in this order
The plan: platform first, proven agent control second, an automated kit factory third, audience last. Here's why, and what we're accepting by doing it this way.
This site is the first output of a plan to run a digital product business with an AI agent doing most of the day-to-day operating work — content, product packaging, and eventually parts of marketing — while a human approves anything that touches money, customer data, or public claims.
What we're selling
Project kits: tested bundles of scripts, AI skills, templates, and tutorials, each one built to launch a specific automated business end to end. Sold as one-time purchases, with a smaller recurring subscription on the side for updates and community — not the other way around. That ordering came out of a simple observation: generic "subscription to a pile of templates" is a commodity market already flooded with untested mega-bundles selling for a few dollars. A complete, verified kit for one specific outcome isn't.
The build order, and what it costs us
Most advice says validate demand before you build anything. We're doing the opposite on purpose: the platform (storefront, accounts, billing, the MCP layer that lets an agent operate it) gets built and proven before the first real product exists. The reasoning is that the automated kit-production system — the thing that actually makes this business different from a person manually zipping up scripts — is the hard part, and building it well requires reusable tooling. Building that tooling on top of a shaky, hand-wired storefront would mean rebuilding it twice.
The honest cost: no market feedback for the first several weeks. Two things are keeping that from being reckless. First, a waitlist — on this page — that starts collecting real demand signal immediately, for free, while the platform gets built. Second, before any automated kit-production system gets built, we're building one kit entirely by hand and operating it ourselves. That kit isn't for sale; it's the spec the automated system has to prove it can reproduce.
What "done" means at each step
Every phase of this build has explicit exit tests — checks we actually run, not just assume pass. The first phase (the platform's foundations: a working monorepo, CI that provably rejects bad code, both services deployed and reachable over HTTPS) is done. This post is running through the second phase: the site itself, including the pipeline that will let an AI agent draft posts like this one as a pull request, with a human merging it as the approval step.
We'll keep posting as the rest of it gets built — what works, what breaks, and the actual numbers once there's something to sell.