Tags: startup, productivity, discuss, ai
Month one: $0. Month two: $0. Month three: $0. Month four: $29. I'm going to be completely transparent about my numbers, because every "how I grew to $10k MRR" thread on here seems to skip the part where the founder ate instant noodles for six months while building something nobody wanted.
My project is an AI automation tool for small businesses — think Zapier but you describe your workflow in plain Chinese or English and it just works. Or at least, that's the pitch. The reality was messier.
Let me walk you through the actual timeline. I quit my corporate job in March 2026 with about eight months of runway saved up. Not a lot. Definitely not the "raised a seed round and have two years to figure it out" story you usually hear. Just a guy with some savings, a ThinkPad, and a questionable amount of confidence. The first version shipped in late April. Looking back, "shipped" is generous. I deployed a Next.js app to Vercel with a single form that called OpenAI's API and crashed if you typed more than 200 characters. My total users: four friends who felt obligated to test it. Their feedback was brutal in the way only friends can be. "It's slow." "I don't get it." "Why would I use this instead of just asking ChatGPT directly?"
May was the rebuilding month. I rewrote the core workflow engine three times. Each time I thought I'd cracked it. Each time I discovered a new class of edge cases that made the previous architecture look like a house of cards. I spent $127 on OpenAI API credits testing my own product. My runway calculator started blinking yellow.
June is when I almost quit. I'd been at this for three months with nothing to show for it. My GitHub looked active, sure, but my bank account was a countdown timer. I applied for exactly one job — a senior frontend role at a startup in Shanghai. Got to the final round. Didn't get it. In retrospect, that rejection might have been the luckiest break of this entire journey, because it forced me to look at my own project again and ask: what would make this worth paying for?
The answer was embarrassingly simple. I wasn't solving a real problem. I was solving the problem I thought people had. The pivot came in late June: instead of a generic workflow builder, I narrowed the focus to e-commerce sellers who needed to automate their customer service across multiple platforms — Taobao, Shopee, Lazada. Specific. Painful. A problem people actually have.
July 12, 2026. First paying customer. $29/month on the early adopter plan. I didn't do a launch. I didn't post on Product Hunt. I found this person in a niche online community for small e-commerce store owners, where I'd been lurking for weeks just listening to people's complaints. Someone mentioned they spent three hours every day copy-pasting customer inquiries between platforms. I reached out directly. Awkward, but it worked.
That $29 hit my Stripe account and I stared at the notification for a solid minute. It wasn't about the money — $29 doesn't even cover my coffee habit. It was proof. A stranger, somewhere out there, found enough value in something I built to exchange real currency for it.
Here's what the numbers actually look like right now: one customer. $29 MRR. Total lifetime costs: roughly $3,400 in living expenses, $280 in API fees, $12/year for the domain. Burn rate: about $1,200/month in a tier-2 Chinese city. Runway remaining: honestly, not enough. I need to hit $1,000 MRR by November or I'll need to start freelancing.
Am I scared? Yes. Is this sustainable? Not yet. Do I regret it? Not even a little. That single customer taught me more about product-market fit than any book, course, or conference ever did. Every conversation since has been grounded in what that first person actually needed, not what I imagined people needed.
The next milestone is ten customers. I have no idea when I'll get there. But I know exactly what I'll do when I do: the same thing I did for the first one. Listen more than I talk. Solve one specific problem well. Charge money from day one.
I write these numbers down — the ugly ones too — at xundao.xin.