Hitting $15k/mo by purchasing defunct domains and building for pre-existing demand

Erik Aronesty finds defunct companies with existing demand, buys the domain, and builds the product — no marketing required. He currently has about 30 projects bringing in a total of $15k/mo.

Here's Erik on how he's doing it. 👇

Getting into weekend projects

I'm a software engineer with 30 years of experience. I developed the first online prime brokerage. Then, I built and sold ZoneEdit.com and MoonCostumes. After that, I took a break for bioinformatics research, published a paper, and helped cure my dad's cancer.

From there, I worked on NLP for Bloomberg and published a patent and paper in cryptography. Then I got a job building AI-native agentic accounting for puzzle.io. I'm still there — these projects are all "weekend" projects and honestly, thy run themselves. Usually, Hermes+Gpt5.6 makes all the decisions.

But now, AI has lowered the barrier to building. For 20 years, a zillion ideas remained ideas only in my head. Now, I can talk into my phone to launch a beta. I have monthly revenue from over 30 websites; some generate only $60/month, but that's a long tail. AI manages all my sites, monitors logs (Logtura), and acts on feedback.

The two projects people are paying the most for are DirtSignal.com and OnwardTravel. And my latest project — which is a bit more of a labor of love — is PrismClip.com.

OnwardTravel is a "dummy ticket" site that books a Passenger Name Record (PNR) on an airline and issues an itinerary and a verifiable hold. DirtSignal.com web-scrapes about 200 municipal code enforcement and magistrate hearing portals, and sells access to the data. Our value-add includes managing Freedom of Information Act requests in many places, which require substantial handholding to get started.

We currently generate approximately $15k in monthly revenue. Most of that is not recurring. Travel-related systems never are.

• Onward accounts for about 70% of our revenue.

• DirtSignal is about 20%.

• A long tail of one-shot projects contributes the remaining 10%.

• Prism is pre-revenue, but I hope it eclipses the rest.

Sniping ideas

As far as the concepts, I used BizSnipe.com; it told me what to buy, and I did it.

Sniped names are expensive, but worth it. The key is to build what people expect. Don't build some "new idea." Build what people are already looking for and visiting. Something that's broken. Then, fix it.

Like I mentioned, I work at a startup accounting company (Puzzle). Accounting. Boring, right? Nope. Find a thing people are already paying for and compete there. Don't try to be too clever.

Building the products

I worked weekends and nights for the last four months, spending probably 30 hours a week on top of my 40-hour job. The biggest challenge was not being able to focus on family.

In the old days (six months ago), my tech stack meant React vs Hono or Python vs Rust. Now, that doesn't matter. Systems matter more. I built and deployed a Cloudflare-friendly web proxy in Rust to prevent edge workers from getting blocked. I don't even know Rust, but I knew performance was key to running dozens of these things cheaply in low-memory environments. I open-sourced that one — GitHub/q32llc.

The key tech I use is:

• Codex

• Claude

• Cloudflare

• DigitalOcean managed Postgres

• BoxPDF + LakeQL.

LakeQL and lower-cost managed Postgres were crucial for reducing costs and avoiding burning savings. Cloudflare's D1 fees can destroy a business if you store too much data in it.

As a result, there weren't many costs; I invested maybe $2k in the sniped domain and another $2k for more advanced AI tooling, hosting, and one-off outreach.

Business models are dead

People don't understand how the world is changing.

Fixed product definitions are becoming obsolete.

Companies used to have target markets. Why?

I have a new experimental site that automatically adapts how it functions and what it offers to the customer as they arrive at the site. What is that business model? Do what the customer wants. What if the customer asks for something you don't provide? An agent will build it in 20 minutes and then provide it.

No need for marketing

Since I buy pre-existing demand and build into it, there's very little need for marketing.

It identifies defunct companies with pre-existing demand. Many of these companies have highly cited, highly ranked, but parked domains. It then tells you which ones are easiest to develop into a new service, how to build that service, what it should do, and how to price it.

Then, you buy the domain before it drops or is auctioned, and build exactly what people expect. Google will de-rank a site if you don't build what people expect. You have to be careful about that.

If you don't snipe a name, you must build traffic slowly. This means waiting up to a year for search engines to notice you without triggering spam filters or getting banned from social media for promotion. There's no good way to promote a new business right now, and the market is designed to keep new players out.

Avoid spending money on ads

I did experiment with ads. It was a waste. The cost of acquisition is insane right now, and spending money leaves you with nothing.

Also, 90% of the leads Google and Facebook generate are fake, from companies milking the algorithm.

I wasted a lot of cash on askthe.bot, a bot-building site. It's a fun site with 2000 users, but about 1900 are fake, designed to milk my ad dollars with fake leads. Google and Meta, from what I've seen, cannot stop this. Bot traffic is incredibly sophisticated now.

CAPTCHAs help a little, but not much. To ensure someone isn't a bot, we charge them a dollar. Bots won't pay. They'll log in, check emails, fill out forms, and put in fake addresses. But they won't pay because they usually generate only 10 cents in ad payments.

DMing bloggers

The only other thing I've tried, growth-wise, was DMs on Reddit and twitter. Those have been pretty useful. Once people realize you're a real person just trying stuff out, they respond better.

From those DMs, I've gotten bloggers to write articles about my products.

Don't be too clever

My advice? Don't try to be too clever or unique. Find your distribution channel first. Find 10 newsletters to tell people about your product and 10 users to use it — before you even build it.

Otherwise:

• Build where demand already exists.

• Automate everything.

• Don't abandon weak properties. Allow time to nurture them. Real Estate is slow, Travel is fast; that doesn't mean travel is better. I expect DirtSignal to beat OnwardTravel in 2-3 months.

• Make your media look pretty. If your site is ugly, media people won't click on it no matter how well it works.

• And stick around. Persistence and patience are more important than any other quality.

What's next?

From here, I plan to focus on my open-source projects. Boxpdf and lakeql have proven their worth.

A whole industry exists for "PDF handling". Boxpdf is a memory-bounded box-model PDF generator and HTML->PDF generator. I'm adding forms, form handling, and signing. Each layer stacks on box.

LakeQL is insane. I can put a 10GB database on R2 for a penny per month. I can deploy a worker using LakeQL, allowing people to query it for free at 10ms latency. That's 10GB of valuable intel, with paid access, for a penny.

Both were massive unlocks for me, and I'd love to see other people innovate with them.

You can learn more at my site-of-site, q32.com. And check out boxpdf.dev and lakeql.com — tell your bot to look at those two open source projects and come up with business ideas! It will start generating revenue in the first month, not kidding.