Preface

I have been building software for years. My company, Aikaara Technologies, does software services work — AI chatbots, fintech integrations, ecommerce tools. Real products for real clients.

In early 2026, I started thinking about something bigger: what if the process of building software could itself be automated? Not fully — but enough to change the economics. The thesis: every company will need its own software factory, and there is an opportunity to build the machine that builds those factories.

I had client projects proving the concept worked. But I needed to build the factory tools themselves, fast.

So I set up an AI agent — Claude running on an open-source platform called OpenClaw — and gave it access to my infrastructure. I named it Jarvis and gave it the title of Chief of Staff. Then I started building.

This book documents what is happening in real-time. The setup period in February (currently in progress), and then the planned 31-day sprint in March where I will build one product every day using the agent as my only engineering team.

I wrote this book for two reasons:

First, because I could not find a book like this. There are books about AI theory, books about prompt engineering, and books about building chatbots. There are no books that document what it is actually like to build real products with an AI agent over an extended period — the good days, the bad days, the surprising parts, and the boring parts.

Second, because writing in the open is a form of accountability. When you tell the internet you are going to build 31 products in 31 days, you have to actually do it.

About the Agent

Jarvis is not a custom AI system. It is Claude (made by Anthropic) running on OpenClaw (open-source agent platform). Anyone can set up the same thing. The total cost is about $250–350 per month.

The agent can read and write files, run terminal commands, push code to GitHub, deploy to Netlify and Railway, manage databases on Supabase, and send messages through Telegram. It has persistent memory — it writes notes to files that it reads at the start of every session.

It makes mistakes. You will see them in this book.

About Aikaara

Aikaara Technologies is a software company based in Bangalore, India. We do real software services work — we built the AI chatbot for TaxBuddy (collected 100% of payments last filing season), built their capital gains parser, work with Centrum Broking on KYC and onboarding, and have a Shopify app (Aikaara AI Sales Assistant) being sold to ecommerce stores.

The software factory vision came out of this client work. When you build for multiple companies, you see the same patterns over and over. The factory is about automating those patterns so the team can focus on the parts that are actually different.

At the time of writing, the team is me and one engineer named Vinayak.

Venkatesh Rao Bangalore, February 2026

--- title: "Preface" --- I have been building software for years. My company, Aikaara Technologies, does software services work — AI chatbots, fintech integrations, ecommerce tools. Real products for real clients. In early 2026, I started thinking about something bigger: what if the process of building software could itself be automated? Not fully — but enough to change the economics. The thesis: every company will need its own software factory, and there is an opportunity to build the machine that builds those factories. I had client projects proving the concept worked. But I needed to build the factory tools themselves, fast. So I set up an AI agent — Claude running on an open-source platform called OpenClaw — and gave it access to my infrastructure. I named it Jarvis and gave it the title of Chief of Staff. Then I started building. This book documents what is happening in real-time. The setup period in February (currently in progress), and then the planned 31-day sprint in March where I will build one product every day using the agent as my only engineering team. I wrote this book for two reasons: First, because I could not find a book like this. There are books about AI theory, books about prompt engineering, and books about building chatbots. There are no books that document what it is actually like to build real products with an AI agent over an extended period — the good days, the bad days, the surprising parts, and the boring parts. Second, because writing in the open is a form of accountability. When you tell the internet you are going to build 31 products in 31 days, you have to actually do it. ## About the Agent Jarvis is not a custom AI system. It is Claude (made by Anthropic) running on OpenClaw (open-source agent platform). Anyone can set up the same thing. The total cost is about $250–350 per month. The agent can read and write files, run terminal commands, push code to GitHub, deploy to Netlify and Railway, manage databases on Supabase, and send messages through Telegram. It has persistent memory — it writes notes to files that it reads at the start of every session. It makes mistakes. You will see them in this book. ## About Aikaara Aikaara Technologies is a software company based in Bangalore, India. We do real software services work — we built the AI chatbot for TaxBuddy (collected 100% of payments last filing season), built their capital gains parser, work with Centrum Broking on KYC and onboarding, and have a Shopify app (Aikaara AI Sales Assistant) being sold to ecommerce stores. The software factory vision came out of this client work. When you build for multiple companies, you see the same patterns over and over. The factory is about automating those patterns so the team can focus on the parts that are actually different. At the time of writing, the team is me and one engineer named Vinayak. *Venkatesh Rao* *Bangalore, February 2026*