From Zero to Production: One Month, $750, and an AI Co-Founder

If you want to go fast, go alone. If you want to go far, bring an AI

- adapted African proverb by Claude Opus 4.5

PS: I found it as extremely funny and deep quote generation

Six Months of Thinking, One Month of Shipping

In May, I discovered Claude Code.

Before that, AI agents felt like a fancy toy - impressive demos, unreliable in real work. After a few days with Claude Code, I've got sbozhed. For the first time, I could see a real opportunity to use AI in my commercial contract work.

The irony was hard to ignore. Everyone at the company talked about AI constantly. Slides, meetings, opinions. Nobody was actually doing anything with it.

So in my free time and holidays, I built a complete feature solo. Two weeks. Shipped.

That moment changed everything for me.

I realized how painfully slow most companies are when it comes to innovation. I've seen this pattern before. Vue 2 to Vue 3 was ready in 2021. Most companies started migrating in 2024 - when it was "finally reliable" with three years of waiting for permission.

I don't want to wait three years to learn what's already here.

So I spent three months thinking about what I actually wanted to build. The answer became clear: something public, something open source, something where every mistake and every win is visible.

I decided to learn in the open - to document every action on sbozh.me and get sbozhed in the process.

Then came two more months of circling the idea. Enjoying unemployment. Letting it crystallize.

December 1st. Official start.

Winter came.

And here you are - reading this.

Enjoy.

The $1,000 Constraint

Honestly, this probably wasn't the wisest decision without a stable source of income.

The result also looks deceptively simple. If you squint, it could pass for a $50 WordPress theme.

Believe me - I know. And believe me - that was intentional.

I wasn't trying to build something impressive. I was building a foundation.

I don't like thinking about resources while I'm building. I'm a builder, not a manager. Optimizing budgets, forecasts, and ROI is a different mode of thinking - there will be a separate persona for that later.

This time, I wanted momentum without negotiation.

So I set a hard limit: $1,000.

The approach was simple. I added $100 at a time. Each top-up was a decision to keep going. Once the money was in, I never looked back at whether it was "worth it." No micro-optimization. No sunk-cost regret.

Was it painful at moments? Yes.

Was it wasteful? Probably, in places.

Was it effective? Absolutely.

What I bought with that money wasn't a website. It was speed, focus, and a real understanding of what it means to ship with AI in production.

And I'm more than happy with the results.

The Numbers

This section is completely AI generated. You can find sources that I entered to a chat session in "Attribution" section at the bottom page. Desktop Claude reviewed my Roadmap, Changelog and commits from public repo.

- The Architect

Here’s what December actually looked like.

The cost

$750 spent on AI tokens. Technically, $744.26, but let’s keep it clean — the round number tells the story.

Lines of code

59,065 lines of code accepted. Not written by me — accepted by me. Claude proposed, I reviewed. 98.4% of suggestions made it through. That’s not blind trust. That’s hundreds of iterations building a working partnership.

Tokens used

506 million tokens in. 2.2 million tokens out. A 227:1 ratio. For every token Claude generated, it consumed 227 tokens of context. Not inefficiency — depth. Understanding before speaking.

115 hours of human time. Roughly. Some days were 10-hour sprints. Some days I touched nothing. The AI doesn’t sleep — but I do.

349 commits. 82 releases. From v0.0.1 to v1.2.0 in 29 days. Most releases were small. Ship early, ship often, fix forward.

One spike tells the story better than any average. December 14th and 15th — $287 combined. 38% of the month’s spend in 48 hours. Those two days took sbozh.me from “almost ready” to production: infrastructure, CI/CD, SSL, monitoring, final bugs. The kind of sprint that used to require a team.

Average daily cost: $29. Some days $3. Some days $177. The rhythm wasn’t constant — it followed the work.

Christmas break: $0. Even builders rest.

Was it worth it? I shipped a production platform with:

• Self-hosted analytics
• CMS and blog system
• CV generator
• Error monitoring
• 90%+ test coverage

Solo. In a month.

You tell me.

Let’s Argue About This

One of my favorite legends is about Diogenes, walking the streets with a lamp, searching for an honest person.

I’m doing something similar - but with builders. I want to argue with people who are creating.

AI is here. What’s your reason not to use it?

“I hate AI because my old world worked fine without this complex auto-complete.”

- The thought the retrograde will never say out loud

You might think it kills creativity. Maybe. But here are my results. What are yours?

I’m not trying to brag. Looking back, what I shipped is essentially a very expensive, neutered WordPress theme - running on Next.js, powered by AI. That’s it. Nothing magical. Just a foundation.

If you have thoughts, critiques, or just want to debate, come to the sbozh.me community of builders. Let’s have an honest conversation. Judge the work, not the fear.

The Work Continues

This year was fire - sometimes literally. I learned a lot. I decided to create even more. And most importantly, I’m committed to keeping the work going.

Right now, it often feels like I’m sending messages into the void. But as they say: stare into the darkness long enough, and the darkness will stare back. I believe the same applies to becoming a public builder and media persona. At some point, the void will answer.

I wish everyone a strong finish to 2025 and all the best in 2026.

Except the rashists - may the void finally hold them accountable for what they started.

Please think about the Heroes and the lost Souls.

See you next year.

— The Founder