How I'm Building a 1M ARR Venture with AI As A Bootstrapped Solo Founder (15K revenue last mo; ~10M last exit w/ Proxycurl)

Background: I'm a seasoned bootstrapper operator and I have built many profitable companies in the past. From a VPN, to Proxycurl which was the largest LinkedIn scraping API I sold last year after settling a lawsuit with LinkedIn. I also write code (before the advent of Claude Code). Also, no AI was used used in writing this article. I'm publishing this post with warts and grammar mistakes as it is.

I intend to share how I'm going to build a 1M ARR company as someone who has done this multiple times, albeit this time as a solo founder with no intention of hiring. I'm determined to make the first 1M with NinjaPear solo. So far, I've been making more than 10K in the last 2 months. So NinjaPear is already profitable :)

First things first, I really don't know what I'm doing so don't take what I say as gospel. Instead, use my story as a sounding board in your repertoire of stories of operators are building solo founder startups. Sure, I have had some startup building experience and they allow me to sidestep mistakes that I have made in the past, and hone in on things that worked out well for me historically. But I'm still struggling. For example, right now, I personally feel like I have not found product market fit with NinjaPear.

I never intended to start another company after Proxycurl's exit because company building is very exhausting. However I fractured my foot in January this year and I was stuck infront of my keyboard and I started to toy around with Claude Code. And lo and behold, I started building out of habit and ended up buiding and deploying the NinjaPear prototype in two weeks. I wrote a blog post and launched it. I was sold that I never needed to hire another (grunt work) dev. Case in point, Proxycurl had around 60 employees with more than half of them as SWEs before I shut it down).

So, NinjaPear is 6 months old, this has been my process on a macro level:

1. Reducing operational expenses to the bare minimum.
2. Iterating non-stop to reach product-market fit with customer feedback.
3. Agentically automating whatever that can be automated.

What's more important are the things that I'm not doing:

1. I am not tokenmaxxing or trying to automate everything. To be specific, I'm not automating taste. There is significant human in the loop, that is myself in decision making.
2. I am not automating marketing. Humans buy from humans, and not AI. (Or why I think Polsia is an amazing spam machine but a shit product).
3. I am not using AI to figure out the product plan or the pitch.
4. Not trying to be a lifestyle influencer/content creator (hah).

Reducing operational expenses

I'll let you in on a little secret. NinjaPear's hosting costs $0 including cost of electricity. Hah, I found an old Intel NUC that I had purchased many years ago that was used to power the signing of transactions with the hot wallet of NuMoney (cryptocurrency exchange startup), wiped it, and plugged into my home internet. And it turns out that I live in Singapore and I have solar panels at home, hence electrical bills are $0!

Compare and contrast it to my time at Proxycurl, for which our monthly spend on hosting on the managed Kubernetes clusters and backups at Digitalocean were $6000-$8000/month. It was ridiculous. The justification was that money spent on DO is far cheaper than hiring engineers to manage backups and orchestrate scaling.

Turns out with Claude Code/Codex, setting up a Kubernetes cluster is trivial. Since then, I have further invested in a Framework desktop, so now I have a cluster of two nodes, all hosted the same way, for $0 ignoring the cost of the home internet I was going to pay for anyways. (Yes, I'm conveniently ignoring the initial capex to purchase these computers).

What do I pay for then?

That said, I still have expenses.

1. I have a 5x max plan for Codex, and a 5x max plan for Claude Code. That's $200/mo.
2. I have some underlying vendors that serves as raw cost for the data enrichment pipeline.
3. Openrouter LLM costs.

But other than these, nothing much. In other words, minimal expenses to keep NinjaPear running.

Non-stop iteration in search of PMF

Based on experience, you will roughly know when you achieved product-market fit because your focus will move from trying to iterate on the product message fit, to having to focus on scaling the product for use. Also, churn goes down and recurring purchases will increase.

I thought NinjaPear had it about two months ago but the revenue this month dipped so I'm not there yet. So if you notice NinjaPear's home page iterating again and again, this is why. I'm still figuring out the definitive answer to "What is NinjaPear". Right now, NinjaPear is a B2B enrichment and competitive intelligence data platform.

As a solo founder in the age of the AI, the product is practically free in the grand scheme of things and it is all about marketing and solving the right problems for your potential customer persona. And iterating means doing this, repeatedly, until you find something that seems to work ok, then hone on it until it works even a bit better, repeatedly.

Agentic Automation

In my previous companies, we have team responsibilities as such:

• Customer success - to provide a point of contact for paying/would-be customers to find out what they like about our product; and to prevent them from churning.
• Customer support - to assist customers with support questions. These include billing.
• Administrative / Bookkeeping - to get your books in order for annual reports, etc.
• Sales team - to close larger Enterprise deals

As a solo founder, some responsibilities/roles disappeared (phew). These usually include managerial role.

Remember how I said that with AI, building is "free" in the grand scheme of things?

Well. What used to take a dedicated headcount as the role of the book-keeper to add receipts/invoices into Xero, is now automated as a Kubernetes deployment.

I don't have a sales team because I don't have PMF yet; can't sell if I haven't got the product messaging figured out.

For customer support, I reused an early version of an AI live chat widget that I had launched before the current rendition of NinjaPear.

For customer success, I built out an agent that sent an AI customised email, something like this:

Hi X,

Thanks for topping up and supporting NinjaPear - I really appreciate it. I noticed you’re building at Y, helping organizations get AI-ready through upskilling, which is a space I find really interesting.

I’d love to understand what problems you’re looking to solve with NinjaPear, and make sure we help you get the most out of it. If you have any questions or need anything, just reply here.

Emails that were sent out based on various action triggers of a customer's journey in our product.

There is a lot more that I have agentically automated at NinjaPear; actions that used to be a full-time position but now automated away. AI that doesn't seek "career growth" or annual bonus payouts. Honestly, I'm happier than ever.

One more thing

This time, I'm spending every cent I earn from NinjaPear and reinvesting it back into the business. Specifically, I'm doing things like boosting tweets/X posts, sponsoring newsletters, Reddit ads. I don't know. I haven't done it yet other than it sponsoring a Japan vacation that I just went for.

I have an unfair advantage this time. For the first-time in my life, I'm able to build a venture for which I have no need for the venture to provide sustenance to myself and my family. This time, I'll never sell NinjaPear. Nor will NinjaPear ever accept an investment :)

I'm going to keep rambling

Obviously, NinjaPear is not anywhere near 1M ARR and I'm going to keep sharing what I'm doing as a solo founder. I'm going to keep writing about my journey and you can subscribe to my 1M ARR Solo Founder journey Newsletter, and follow me on X!

I intend to share a story every work day. Trying to work my writing muscle again. My next article will be about what I'm not automating. And why I think having myself in the loop matters.