I run 2 companies on a $5 AI stack. Here's exactly how.

No $200 ChatGPT Plus. No enterprise contracts. Just n8n, DeepSeek, and open-source tools that actually work.

Get the stack breakdown

Followed by 0 founders... for now. Join from Day 1.

my-ai-stack
$ cat workflow.json
"trigger": "webhook",
"model": "deepseek-v3",
"provider": "openrouter",
"cost_per_1M": 0.50,
"monthly_spend": 4.78
stack ▶ $0.14 spent today
84%
tasks automated today
Scroll
Origin Story
"I got banned from ChatGPT. It cost me $147. That ban saved my business."

I was running AI workflows on ChatGPT Plus — $20/mo, reliable, simple. Then one day, OpenAI flagged my account. Banned. No warning. No appeal process that made sense. $147 in API credits locked behind a deactivated account.

That forced me to find alternatives. I discovered n8n, learned to self-host on a $3.50 VPS, and switched to DeepSeek V3 via OpenRouter at $0.50 per million tokens. My costs dropped from $200/mo to under $5.

Now I run two companies — Authentic Education and Botsquash — from Bangladesh and Cyprus, entirely on open-source AI tools.


I document everything so you don't waste money on AI that doesn't deliver.


What I Build

No theory. Just what works.

Every piece of content comes from running real businesses with real AI.

Real Workflows

Complete n8n automations with screenshots and actual code. No diagrams — real JSON exports you can import.

See example

Cost Breakdowns

Exact dollar amounts. Self-hosted vs SaaS pricing compared line-by-line so you know where every cent goes.

See example

Operator Lessons

Running AI-automated businesses day-to-day. What breaks, what scales, and what I'd do differently.

See example

The Stack

The $4.78/mo AI stack

Click any tool to see exactly how I use it and what it costs.

n8n
Self-hosted automation engine

200+ integrations. Runs on a $3.50 Hetzner VPS. Zero vendor lock-in. Full control over your workflows. I run 47 active automations across both companies from a single instance.

DeepSeek V3
OpenRouter, $0.50/1M tokens

Outperforms GPT-4 on coding and reasoning benchmarks. Costs 40x less than GPT-4 Turbo. No rate limits on OpenRouter. I process ~2.4M tokens per month across all workflows.

Jina AI
Web search & embeddings

REST API for web search, content extraction, and embeddings. I use 0 paid APIs — Jina's free tier covers all web scraping and search needs for both companies with room to spare.

Open WebUI
Chat interface, self-hosted

Clean ChatGPT-style UI that connects to any LLM via API. Self-hosted on the same VPS. No data leaves my server. I use it for ad-hoc queries, brainstorming, and quick prototyping.

Get the $5 AI stack breakdown

Plus weekly workflows I actually use. No theory. No fluff.

No spam. Unsubscribe anytime. I hate clutter too.