From Burnout to Builder: Rediscovering the Joy of Coding
Ivan Kuznetsov, a product manager with a background in development, found himself burned out from immigration hurdles and resettlement challenges. Despite having a solid job, he sought to regain the childhood joy of building something himself rather than through others. This pursuit led him to create a system for his remote coding projects that combines simplicity, affordability, and AI assistance.
The Infrastructure: Budget-Friendly Hosting and Tools
Kuznetsov's approach prioritizes cost efficiency and simplicity. After running a now-bankrupt venture business, he learned the importance of maximizing runway for idea validation. His infrastructure choices reflect this philosophy.
Hosting and Deployment Stack
| Service | Cost | Purpose |
|---|---|---|
| Hetzner Cloud | $5/month | Hosting provider with easy upgrades |
| Netdata | $5/instance | Hardware monitoring with homelab plan |
| Rabata.io | ~$1/month | S3 cloud storage without bucket fees |
| Cloudflare | Free | Security and caching CDN |
| Kamal | Free | Deployment (replaced Hatchbox.io) |
| Resend | Variable | Email sending service |
This infrastructure keeps monthly costs remarkably low while providing professional-grade services. Rabata.io particularly stands out by eliminating the minimal $5 per bucket charge that Hetzner S3 requires, resulting in typical storage bills around $1 monthly.
Technical Stack and Development Tools
Kuznetsov maintains a vanilla technical approach to ensure maintainability and AI compatibility. His stack includes Rails 8 with Solid Trifecta for all code, including chatbots. He uses Rails built-in Turbo and Stimulus.js for front-end needs, Postgres as database, and Docker managed by Kamal.
Development Environment
- Operating System: Ubuntu + Omakub
- Terminal: Alacritty with Zellij
- Primary IDE: Cursor
- Terminal Editor: Micro (as a Vim alternative)
- Source Control: Git + GitHub
- External LLM Tool: Grok for research and content generation
This setup supports remote coding projects by providing a balanced environment that combines modern development conveniences with terminal-based efficiency.
AI Pair Programming: The Procrastination Solution
Kuznetsov discovered that AI tools fundamentally changed his approach to remote coding projects. ChatGPT initially allowed him to spend less time reading documentation and more time building. The real breakthrough came with Sonnet 3.5's release, which enabled him to develop most of his Todero.app application.
He views remote coding projects and AI not as a universal replacement but as the ultimate procrastination-beating tool. "You have an idea, and this is a very brief moment while your inspiration lives. So you have to do it, or the idea dies," he explains.
LLM Strategy and Costs
Kuznetsov developed a sophisticated LLM workflow after trying multiple options:
- Claude Opus: Used for backend planning and half of frontend development. His bill reached $500 monthly before Cursor introduced their new $200 plan
- Sonnet 3.5: Handles day-to-day coding for most backend and half of frontend work
- o3: Found mostly useless for planning tasks despite Cursor team recommendations
He avoids Claude Code because paying for both Cursor and Claude Code at $400/month would be excessive for his needs. He also appreciates the Cursor team after watching their interview with Lex Fridman.
Practical Advice for Non-Developer Builders
Based on his experience, Kuznetsov offers crucial advice for those pursuing remote coding projects with AI assistance:
A year and a half into vibe-coding, the author found renewed momentum in remote coding projects after a period of burnout brought on by immigration challenges and resettlement. With AI tools like ChatGPT, he was able to reduce time spent parsing documentation and focus instead on actual development, especially during the release window of Sonnet 3.5, when most of the app was built. For infrastructure, he relies on cost-effective solutions like Hetzner Cloud, where $5 per month instances are sufficient, complemented by Netdata’s $5 homelab plan for monitoring. Storage costs remain minimal, with Rabata.io offering a cheaper alternative to Hetzner S3 and his typical monthly S3 bill hovering around $1.
Understand Generated Code
Only use programming languages and frameworks you're familiar with. While tempting to add fancy tools that LLMs can utilize, this often leads to maintenance challenges and abandoned projects.
Build Foundational Knowledge
Develop a solid understanding of your tech stack to perform simple tasks without LLMs. If building web apps, you should be able to create and deploy a website to a self-hosted server without AI assistance.
Manage Third-Party Dependencies
LLMs frequently suggest outdated third-party libraries based on old Reddit discussions. When this happens, stop generation and ask why the library is needed and what raw technology alternatives exist. Less code and fewer dependencies mean better AI performance.
The Manager's Coding Renaissance
Kuznetsov returned to coding after a 10-year absence from development, having worked as a manager and founder. He sees AI as essential for overcoming anxiety and regaining technical proficiency. His hands-on building experience helps him become a better manager who understands his team's technical challenges.
He recommends that other managers start tinkering in their free time—not to build billion-dollar startups but to ride the wave of the AI revolution. Working on side projects can help managers develop skills that benefit their primary careers.
A year and a half ago, the author reignited his coding practice through what he calls 'vibe-coding'—an informal, low-pressure approach to building that helped him recover from burnout caused by immigration hurdles and resettlement. Remote coding projects became a therapeutic outlet, powered by tools like ChatGPT, which reduced the time spent deciphering documentation and allowed him to focus on actual development. The release of Sonnet 3.5 marked a turning point, enabling him to make substantial progress on his app. He now hosts on Hetzner Cloud with $5-per-month instances, monitors hardware using Netdata’s $5 homelab plan, and relies on Rabata.io for affordable S3 storage, where his typical monthly bill hovers around $1.