My $0 to First Customer: The Ugly Numbers
The project started as a general-purpose AI workflow builder—essentially a "plain English/Chinese to automation" tool. I quit my corporate job in March 2026 with only eight months of runway and a ThinkPad. By late April, I had a Next.js app deployed on Vercel. It was a disaster; the OpenAI API calls crashed if a user entered more than 200 characters. The only "users" I had were four friends whose feedback was a cold shower: they didn't get the value proposition and wondered why they shouldn't just use ChatGPT directly.
May was a cycle of failure. I rewrote the core engine three times, only to find new edge cases that collapsed my architecture. I was burning API credits and my savings, but I was building in a vacuum. By June, I was desperate enough to apply for a senior frontend role in Shanghai. I made it to the final round and failed.
That rejection forced a pivot. I realized I was solving a problem I imagined existed rather than a real-world pain point. I stopped trying to build a "general tool" and pivoted to a specific LLM agent workflow for e-commerce sellers managing customer service across platforms like Shopee and Lazada.
On July 12, 2026, I landed my first paying customer: $29/month. No Product Hunt launch, no viral thread—just direct outreach in a niche community where I'd spent weeks listening to sellers complain about manual copy-pasting.
The current balance sheet is brutal:
- MRR: $29
- Lifetime API Costs: $280
- Living Expenses: ~$3,400
- Domain: $12/year
- Monthly Burn: ~$1,200
I'm operating on a razor-thin margin and need to hit $1,000 MRR by November or return to freelancing. However, that one stranger paying for the product provided a more honest lesson in product-market fit than any course ever could. I'm no longer guessing; I'm building based on actual user friction.