Every product, taken apart.

FormCraft lets non-technical teams build embeddable forms, collect responses, and route leads to email or Slack — without writing a line of code. Built for the operators who are tired of paying $99/mo for Typeform when they only need 5 forms.

The bet

That a long tail of small teams would happily pay $19/mo for "Typeform, but only the 5 forms I actually use." Most form builders price for the enterprise; almost nobody prices for the operator with a clipboard.

What worked

Brutal scope discipline. FormCraft does five things and refuses the sixth. That made it cheap to run, easy to explain, and impossible to out-feature on the things its users actually cared about. Organic SEO on "Typeform alternative" did the rest.

What broke

Churn from seasonal businesses I never solved cleanly. A landscaper needs forms in spring and cancels in winter. Annual plans helped, but I should have built a "pause" tier two years before I did.

PingAlert checks your URLs every 60 seconds and fires an SMS or Slack message the moment something goes down. No dashboard to babysit — just quiet until it's not. Intentionally simple: one screen, one plan, $9/mo.

The bet

That "boring and quiet" beats "powerful and busy" for uptime monitoring. The incumbents sell dashboards; I bet most people just want a text message when the site is down and silence otherwise.

What worked

One screen, one plan, $9/mo. No free tier to support, no pricing-page paralysis. The simplicity was the product — support tickets are nearly zero because there is almost nothing to misconfigure.

What broke

SMS costs scale with success in a way SaaS margins do not love. A handful of customers monitoring hundreds of endpoints quietly became unprofitable before I noticed. Usage caps now exist; they should have shipped on day one.

DocuFlow is a REST API that accepts HTML or a template ID and returns a rendered PDF in under 2 seconds. Built for developers who are tired of wrestling with headless Chromium in production. Pay-per-use, no monthly minimums.

The bet

That developers would rather pay per PDF than run headless Chromium in production ever again. Anyone who has tried to keep Puppeteer alive under load knows the pain I was pricing against.

What worked

Pay-per-use with no monthly minimum removed the only objection that mattered. Developers could try it on a side project for cents and expand without a procurement conversation. The API surface stayed tiny on purpose.

What broke

Cold-start latency on Lambda was a constant fight, and a "render this 80-page invoice" request could blow the timeout. I under-invested in a queue for the heavy jobs and lost a few high-value customers to it.

ListPilot cleaned and validated email lists to reduce bounce rates before sending campaigns. Launched into a market that already had three well-funded competitors. Couldn't differentiate on price or features — turned off the servers in December 2022.

The bet

That I could win an email-validation market on better UX. The bet was wrong before the first commit: three funded competitors already owned the keywords, the integrations, and the trust.

What worked

Honestly, the shutdown. Killing it fast — before sinking another year in — was the most profitable decision of the whole project. The code became boilerplate I reused elsewhere; the lesson became the filter for every product since.

What broke

Everything upstream of the code. I validated that the problem was real and never checked whether the *market* was winnable. Real problem, unwinnable market, is still a no.

Lesson: I solved a real problem in a market I didn't understand. Validate the competitive landscape before you write a single line of code.

ShipKit was an opinionated Astro + Tailwind starter with auth, billing, and a dashboard shell pre-built. Sold to a developer tools studio in October 2023 who wanted to build on the foundation rather than maintain their own.

The bet

That the boilerplate I kept rebuilding for my own launches was worth selling to other builders. If I needed auth + billing + a dashboard shell every time, surely others did too.

What worked

It did sell — and the acquirer wanted the foundation, not a flip. Productising my own internal tooling turned sunk cost into runway. The cleanest eight months of runway I ever bought.

What broke

Maintaining a boilerplate is a second job: every dependency bump is a support ticket waiting to happen. Selling it was the right call precisely because I did not want that job. Know when an asset is really a liability.