Steve Hanov keeps getting rejected from pitch nights. Not because his products don't work. He has actual users and real MRR. The problem is that his tech stack costs twenty bucks a month, and investors can't figure out why he'd need their money.
Hanov, based in Waterloo, Ontario, runs multiple SaaS products including Zwibbler.com and a financial research tool called eh-trade.ca. His philosophy is straightforward: a $20 monthly burn rate gives you the same runway as a million-dollar raise with a bloated startup, minus the stress and the board meetings.
His setup sounds almost comically lean. A single VPS from Linode or DigitalOcean for five to ten bucks. Go backends compiled into static binaries and scp'd to the server. SQLite with Write-Ahead Logging instead of a managed Postgres instance. No Kubernetes, no container orchestration, no AWS console designed to extract billing upgrades. One server means you know exactly where the logs live and exactly why it crashed.
Where it gets interesting for AI builders is his local inference setup. Hanov runs local AI servers like Lemonade on a used RTX 3090 he grabbed off Facebook Marketplace for $900. He built his own agentic research tool called laconic, optimized for tight 8K context windows, which pages conversation history in and out like virtual memory. For batch tasks like analyzing 11,000 stocks, local inference means no API bills and no panic when a prompt bug forces a complete rerun. This approach differs from relying on local AI agents like Gemma Gem that operate entirely within the browser.
His coding workflow is equally cheap. Hanov uses GitHub Copilot and exploits the fact that Microsoft charges per request, not per token. He writes detailed prompts, tells the agent to keep fixing errors, and walks away. His monthly AI bill stays around sixty dollars.
The Hacker News crowd had mixed reactions. Some called the advice obvious. Others confirmed that SQLite with WAL is a genuine money saver and pointed to Hetzner as an even cheaper hosting option. But the core point stands: before you scale, you have to survive. And surviving is easier when your infrastructure costs less than a Netflix subscription.