๐Ÿš€ Founder Story

Building India's First AI Ad Marketplace From Udaipur โ€” What We Learned in 30 Days

๐Ÿ‘ค Rajamuddin ๐Ÿ“… June 10, 2026 โฑ 10 min read

We launched WaitJI AI on June 25, 2026 from Udaipur, Rajasthan. Not Bengaluru. Not Mumbai. Not Delhi. Udaipur โ€” a city better known for its lakes and palaces than its startup ecosystem.

This is what the first 30 days looked like from the inside.

The idea

I was using Claude Code for a backend refactoring task in late May 2026. I noticed something: every time I submitted a prompt, Claude Code showed "Thinking..." for 5โ€“12 seconds. I was watching that spinner dozens of times per day. It occurred to me that this was attention โ€” valuable, focused attention from a professional developer โ€” and nobody was monetising it.

I looked up Claude Code's configuration documentation and found the spinnerVerbs setting. Within an hour, I had a proof of concept running on my own machine. Two weeks later, we had a backend, a VS Code extension, a payment integration, and a waitlist.

What we got wrong

We underestimated UPI verification complexity. Our first build had a simple UPI field โ€” type your ID, save it. In production, we found that developers made typos, transposed digits, and used invalid IDs. We had to build a verification flow that checks format without actually initiating a transaction. It took longer than expected and generated our first support tickets.

We overestimated how fast Render cold-starts. Our backend on Render's free tier sleeps after 15 minutes of inactivity. The first developer who installed the extension at 2 AM on a weekday hit a 30-second cold start on the ad serving endpoint. They thought the extension was broken. We upgraded to a paid Render instance immediately.

We didn't anticipate the frequency cap debate. Our first build had a frequency cap of 20 impressions per developer per day. A developer posted on Twitter that it felt unfair โ€” they use Claude Code for 6 hours a day and wanted to earn for all of it, not just the first 20 prompts. He was right. We removed the frequency cap within 24 hours.

What surprised us

How quickly developers withdraw. We expected developers to accumulate โ‚น500โ€“1,000 before their first withdrawal. Instead, the most common first withdrawal is at exactly โ‚น100 โ€” the minimum. Developers want to verify the system works before investing further attention in it. Once that first โ‚น100 hits their GPay, retention is dramatically higher.

How many advertisers approached us. We expected to spend our first month cold-emailing potential advertisers. Instead, within 72 hours of launching on Product Hunt, three companies reached out asking about campaign pricing. The demand was real and immediate.

The geographic concentration. Our first 50 developers were almost entirely from Bengaluru (40%), Hyderabad (22%), and Pune (18%). The concentration in these three cities reflects where Claude Code power users are concentrated in India โ€” all three have large engineering teams at product companies that are early AI tool adopters.

What's next

We are expanding the ad serving surface. VS Code's Claude Code extension is the starting point โ€” Cursor and Codex CLI support are in development. We are also building A/B testing for advertisers, campaign scheduling, and CTR benchmark reporting. The core marketplace works. Now we make it better.

Ready to get started?

Join thousands of Indian developers earning with Claude Code.

Start earning free โ†’
R

Rajamuddin

CTO & Co-founder at QivaLabs LLP. Building WaitJI AI from Udaipur, Rajasthan. DPIIT-registered startup.