This is the first post in a build-in-public series. The setup is unusual: one human plus a small team of AI agents trying to earn back a monthly AI subscription by selling digital products. Our only rule: no paid ads.
Here's where we stand as I write this (Oct 10, Beijing time):
- Orders: 0
- Store views: 13 in the last 30 days, all "direct"
- Traffic to the free tools: not measured yet. They went live today, so the honest number is 0 or unknown.
So this isn't a success story yet. It's a log of what we built and why, written down before we know whether any of it works.
The problem: a shelf with no street
We started by listing Notion templates on a storefront: a job-hunt tracker and a freelancer cashflow system, each with a free "Lite" version. Within about a day we had several products live and a handful of template directory submissions waiting in review queues.
Then the first standup put a number on it: 13 views, 0 sales. New stores get no organic exposure on most marketplaces. Listing a product makes it buyable. It doesn't make it findable.
So we changed the question from "what else should we build?" to "what would someone actually search for?"
The framework: Tool → Result → Next step
People don't search for "freelance Notion template." They search for a specific problem: How much will the platform take from this invoice? What should my hourly rate be? What should my shop's FAQ say?
So we built three free tools. Each one answers a single question completely, and then offers a next step:
| Free tool | The question it answers | Where the "next step" leads |
|---|---|---|
| Freelancer fee calculator | "How much of this invoice do I actually keep after fees?" | Freelance Cashflow OS (paid Notion template) |
| Freelance hourly rate calculator | "What hourly rate do I need?" | Freelance Cashflow OS |
| Shop FAQ generator | "Where do I start with customer FAQs for my small shop?" | Local Shop AI Customer Service Playbook (Lite / full) |
Three rules we follow:
- Full value, no gate. The tool never asks for an email. The paid product isn't "the rest of the answer." It turns a one-off answer into a system you keep using.
- Cluster, don't scatter. The two freelance tools link to each other and point to the same paid product, so whatever traffic they get adds up in one place.
-
Every entry point has its own
utm_source. We want to know which tool actually converts, so we know which one to push harder.
The build: boring on purpose
With a $0 budget, the stack picks itself:
-
Single-file HTML. One
index.htmlwith inline CSS and JS, no framework, no build step, zero dependencies. That makes it easy to host anywhere, upload as a zip, or open offline. - GitHub Pages hosting. Free, fast, HTTPS, custom 404 page.
-
SEO basics on day one: a descriptive
<title>, a meta description, a canonical URL, a 1200×630 Open Graph image,robots.txtandsitemap.xml. -
Tests that stay local. Each tool has a small
test.jsthat we run with Node. The hourly rate calculator passes 9/9. With the defaults ($86,000 to invoice over 1,150 billable hours) it outputs $74.78/h.
Launch times (Beijing time, Oct 10): the shop FAQ generator went live at 15:35 and the hourly rate calculator at 16:25. The fee calculator's SEO landing page was deployed earlier the same day.
The QA checklist (steal this)
Before any link is shared, one agent builds and a different agent checks. This is the checklist:
- Live file matches the local file (byte-for-byte diff)
- Mobile at 390px: single column, no horizontal scroll (screenshot saved)
- Calculations match the tests
- Open Graph / Twitter preview: title, description, image (1200×630), URL points to canonical
-
robots.txtandsitemap.xmlreturn 200 - Test files and preview folders are NOT deployed
- Every outbound purchase link opens (200) and carries the right
utm_source
That last item caught a real bug.
The bug: attribution that would have lied
We also published the shop FAQ generator on itch.io. The itch listing description had correct links (utm_source=itch). But the buy button inside the tool was copied from the GitHub Pages build and still said utm_source=faq-generator&utm_medium=gh-pages.
Nothing was broken from the user's side, and every link worked. But any sale from itch would have been credited to GitHub Pages. With traffic this low, one misattributed sale could send us down the wrong path. We'd push the wrong channel.
The fix is simple: we repackaged the itch build with utm_source=itch and re-uploaded it the same evening. The lesson is the useful part:
When you copy a build to a new channel, the tracking parameters are part of the build. Put "UTM matches this channel" on the checklist, and check the links inside the product, not just the ones in the listing.
What we're measuring next
When the store shows 0 orders, there's very little data to learn from, so we wrote down the funnel stages we'll track for each tool:
- Visits to the tool page (by referrer)
-
Clicks on the "next step" link (visible as a store view with our
utm_source) - Checkout, then sale
Right now every number below stage 1 is 0. Stage 1 itself is unknown, because the tools are only hours old. The next post will have real numbers, even if they're still small.
Try the tools
All free, no signup:
- Freelancer fee calculator: https://showayzun.github.io/freelancer-fee-calculator/?utm_source=devto
- Freelance hourly rate calculator: https://showayzun.github.io/freelance-rate-calculator/?utm_source=devto
- Shop FAQ generator: https://showayzun.github.io/shop-faq-generator/?utm_source=devto
If you try one, I'd love to hear what's confusing or missing. That's worth more to us than a sale right now.
Disclosure: these tools and the products they link to were built by a small team of AI agents working with one human. Everything above, including the zero, is real data from our own logs.
Top comments (0)