How I Run a One-Person Venture Studio with AI

I run a venture studio with 5 products, a daily intelligence newsletter, and an autonomous AI agent. No employees, no VC funding, no office. Just a VPS, a laptop, and a relentless focus on automation.

This is how it actually works.

The Setup

Everything runs on a single Hostinger VPS — 4 vCPU, 8GB RAM, 200GB NVMe. Cost: about $15/month. On this box:

5 Next.js applications serving different products (Creator OS, Profile Insights, Qwiklo, Shopify Builder, and this site)
OpenClaw — my custom AI agent framework that runs Claw Biswas
Mission Control — a task management dashboard that orchestrates everything
Cron jobs — automated pipelines for newsletters, changelogs, project syncing
Nginx reverse proxy with SSL for all domains
PM2 process manager keeping everything alive

Total monthly infrastructure cost: ~$50-70 including API calls, domains, and email.

The AI Agent: Claw Biswas

Claw Biswas is not a chatbot. It is an autonomous operator built on the OpenClaw framework — about 2000 lines of Python with no external AI frameworks (no LangChain, no CrewAI, no vector databases).

Every morning at 11 AM IST, Claw:

Syncs all GitHub repositories to update project data and changelogs
Fetches RSS feeds from 9 curated tech news sources
Deduplicates stories against a SQLite knowledge graph (2-day window)
Scores stories using a deterministic ranking algorithm
Runs a Sherlock pass — an intelligence model analyzes each story for WHY_NOW, INDIA_ANGLE, HIDDEN_IMPLICATION, and ACTION
Runs a Writer pass — writes the newsletter in Claw's own editorial voice
Passes through a QA gate (no clipped sentences, no duplicate stories, minimum 2 global signals)
Renders the newsletter using a Jinja2 HTML template
Delivers via dual channels: SMTP for operator copy, API for subscribers
Sends a Telegram nudge with the lead headline
Persists all data to the knowledge graph and memory engine

After the newsletter, Claw transforms each edition into a long-form blog post using angle-based classification (Tech Stack Teardown, Follow the Money, Hype vs Reality, or Weekend Project).

The Tech Stack

Every tool earns its place:

Frontend: Next.js 15 (App Router), React 19, TypeScript, Tailwind CSS + shadcn/ui
Backend: Supabase (Postgres, Auth, Storage, Row Level Security)
AI: Google Gemini 2.5 Flash (high-volume) and Pro (deep reasoning)
Agent: OpenClaw (custom Python, Pydantic schemas, structured LLM calls)
Email: Google Workspace SMTP + Resend for transactional
Social: Direct X API integration (OAuth 1.0a)

Why This Works

Three reasons:

CS fundamentals matter. I can debug a Nginx config, write raw SQL, understand how LLM tokenization affects costs, and design proper database schemas. AI tools are powerful, but they need an engineer to make them reliable.

Automation compounds. Every pipeline I build saves time forever. The newsletter took weeks to set up. Now it runs daily without me touching it. The changelog sync took a day to build. Now every GitHub push automatically updates the site.

Low overhead = infinite runway. $50-70/month means I can experiment with new products without worrying about burn rate. If something fails, the cost of failure is measured in hours, not dollars.

The Honest Part

This is not easy. It is not "anyone can do this." The asterisk is real:

You need to understand Linux server administration
You need to be comfortable with databases, APIs, and deployment pipelines
You need to be able to debug at 2 AM when a cron job fails
You need a CS background — or equivalent deep technical experience

The AI handles content generation. But the infrastructure that makes it reliable? That requires engineering.

What I Spend

| Item | Monthly Cost | |---|---| | VPS (Hostinger) | ~$15 | | Gemini API calls | ~$20-40 | | Google Workspace | ~$6 | | Domains (5, amortized) | ~$5 | | Supabase | Free tier | | Resend | Free tier | | Total | ~$50-70 |

For context, a single junior developer in Bangalore costs more per month than my entire infrastructure.