Four rules for my next company

I’ve built two products that people used but didn’t urgently need. Kaguya, a home for visual novel readers, has over 4,200 sign-ups and no revenue. Nila, an AI accountant for Indian CA firms, worked on real bills and no firm paid for it.

Both times I built first. Kaguya took nine months to a basic version, with a designer and a developer I’d hired. Nila took 20 weeks, alone, with coding agents. The agents made me faster, but the pattern was the same.

I don’t think of a startup as an experiment that proves an idea right or wrong. It’s more like sharpening your picture of reality until you’re making something people want. You start with an idea and keep discovering which workflow comes first, which wedge, and who you’re actually selling to. X.com didn’t know online payments would turn out to be the thing. So you accept from the start that the idea might not work out, and you keep discovering anyway. These rules are about making that discovery faster.

1. Find a cofounder first

Startups are a search in the dark. You’ll get five or ten big decisions wrong, and what decides whether you make it is how fast you notice and correct them. I underestimated that. I thought that if I worked hard enough I could make it on my own, and I put too much weight on my own effort. Even if you watch yourself closely, correcting yourself alone is slow. You keep wondering whether you’re doing the right thing, when what you need is someone looking at the same evidence who can say, “This isn’t working, we need to change something.”

With Nila, it took me about two months just to settle on GST as the place to start, and on sales bills within that, and I went through a lot of design iterations that someone working with me every day would have cut short. Most startups that succeed have cofounders. The ones that don’t are exceptions, like Sara Blakely, who’d spent seven years in door-to-door sales before she started. And it doesn’t need to be a perfect cofounder, just a good one you click with. Musk started with his brother Kimball.

So next time, a cofounder is a requirement. If I can’t find the right person, I don’t think it’s worth starting.

2. Understand one person’s problem before building or selling

Both my ideas made sense. Someone was going to build a better Goodreads, and AI was obviously going to change accounting. But “this makes sense” isn’t the same as “someone needs this now.” Kaguya’s readers used it, but none of them would have been in trouble without it.

So the first job is to get in front of one person who has the problem and understand it in their own words. Don’t lead them, and don’t pitch yet. I read Hormozi’s offers and loved them. At Gym Launch he’d fly out on his own money, get a gym its leads and members, and only take a cut of the ones that converted. Nobody says no to that. But an offer like that works once your product already works. For your first customers it’s still a big ask: you’re new, you have no credibility, and they can always say no. I enjoyed the selling side too, handling objections, finding the gap, closing. All of that only works once you have a product that does the job, and with Nila I never got there.

Next time I won’t write any code until I’ve found people who are already trying to solve the problem, and they’ve told me so.

3. Pick a first customer with urgent pain and money

For Nila, I picked CA firms near me on Google Maps, because I could walk in. Most were small, did a modest amount of GST work with two to five people, and bought software the way you buy furniture: once, then keep it for ten years. Even a happy one might have paid ₹1,000 a month.

Now I’d look for one of two things: someone with an unusual, urgent problem who can’t wait, or an organisation big enough to have both the volume and the curiosity to try something new. Legora built its product inside a large law firm whose lawyers used it every day. I never got a single big Indian firm to talk to me, and that was a signal I should have taken seriously.

4. Know how you’ll reach people before you build

I hoped that once a few firms used Nila, I could find distribution later, maybe through Tally partners who already sell into these offices. When I looked at how a successful competitor actually sold, it was a sales team making relentless calls, plus influencer deals. I hadn’t planned for anything like that.

I also noticed that when selling got hard, I built. I put a lot of effort into features nobody had asked for, like business invoicing that other apps give away free, because building felt like progress. Next time, if I catch myself building to avoid selling, that’s the signal to stop and go talk to someone.

What’s next

I used to think intelligence and hard work were what won. Now I’d rank it differently: the market first, then the people you work with, and only then your own intelligence and effort. I’d carried over the habits that worked for me in college into a game where they matter much less. Marc Andreessen says the market matters more than the team, and the team more than the product. I’d read that for years. After Kaguya and Nila, I finally understand it.

I’m now looking at automated manufacturing, something like a foundry where you send us a design and we make it much faster, built around robotic assembly. This time I’m starting with the two things I skipped before: a cofounder, and proof that someone needs it.