After 6 years of unfinished projects, I finally shipped something (by validating with a fake Stripe checkout)

Source: Dev.to

The Problem with Estimates

The last two years I spent trying to figure out what was wrong with me. Why couldn’t I finish anything?

Two and a half months ago everything changed.

I took a freelance project with hourly billing and immediately remembered how much I hate explaining dev costs to clients:

• “Why is this so expensive?”
• “Why does it take so long?”
• “Can you break down every single feature by hours?”

Estimating projects for six years has always stressed me out:

• Estimate too high → lose the client
• Estimate too low → work weekends for free
• Guess right → you got lucky

I started using ChatGPT and Claude to help me estimate, but it was a mess—I had 15 different chats, Obsidian pages everywhere, and struggled to keep context straight. Then I tried creating AI agents with Claude Code (Flutter agent, UX/UI agent, Risk Manager) and stored everything in Markdown files.

The Turning Point

While doing all this, I randomly met a guy at a coworking space. He’d launched 10+ projects and, after listening to my problem, said:

“Dude. Make a landing page. Put a fake Stripe on it. Run ads. See if anyone tries to buy.”

That blew my mind. Instead of spending months studying the market, I could test right now whether people wanted to buy.

Building the Fake Checkout

That evening I knew what to do. I spent a week building a landing page with AI help, added Stripe in dev mode, set up screen recording, bought a domain, and ran Google Ads targeting India (a big dev audience with cheap clicks).

Key trick: when someone went through checkout and clicked “pay,” they were sent to a page saying:

“No money was charged, thanks bro, you’ll be first when I launch.”

Results

After two weeks, 11 people went through the fake checkout. From that moment I never doubted myself, never thought about quitting, and never wanted to try something else.

The product vision became clear: a backlog tool with AI agents that interview you about your project, ask clarifying questions, and give accurate estimates.

Building EstimateFast

I spent two months building it. Instead of spending a day on a client estimate, I built a tool to do it for me forever.

During development I fought the urge to add a thousand features. One of my favorites—used personally at my main BigTech job—is loading a huge GitHub repo and Confluence docs into a RAG/vector database. Now, when my manager dumps tasks on me, I throw them into the tool and get estimates in seconds.

Note: This feature is currently available only locally; I need to focus on promoting what already exists first.

Conclusion & Next Steps

And that’s the story of how I went from six years of abandoned projects to finally shipping EstimateFast.

I have no idea what I’m doing with marketing or acquiring users right now, but at least this time I didn’t quit.

Any advice for a first‑time launcher? What should I do next?

I’m documenting everything on Twitter @pavel_bekoev—the wins, failures, and lessons as a first‑time founder. Would love to connect with other builders!