Hey Jarvis: Your Machine. Your AI. Your Rules. - PDF Edition

Every morning you wake up and immediately start working. Check the markets. Open email. Pull up the news. Scan your portfolio. By the time you finish your first coffee you've visited seven websites and absorbed maybe a third of what you read, because you're doing it all half-awake and simultaneously.

Hey Jarvis is the story of how one engineer got fed up with being his own bottleneck — and built the assistant he couldn't buy.

Jarvis is a fully voice-controlled AI assistant that runs on your Windows PC. It listens for your wake word, understands what you say, thinks with Claude, and answers in a natural spoken voice. It monitors your inbox, scans for contract opportunities, watches the markets, learns your habits, and overlays its responses on your screen without interrupting what you're doing.

It is yours — not a subscription, not a cloud product, not a consumer device with guardrails. Your machine. Your AI. Your rules.

What you build

The complete system, from an empty folder to a working assistant:

• The voice loop — wake word detection with a custom OpenWakeWord model, voice capture with silence detection, speech-to-text through faster-whisper, and spoken responses through Kokoro TTS
• The brain — Claude wired in as the reasoning layer, with intent routing and conversation context
• The plasma overlay — an always-on HUD that displays responses without stealing focus from what you're doing
• The trading dashboard — grid bots, an Alpaca stock bot running RSI(2) mean reversion, and the backtesting tool that tells you whether a strategy has an edge before you trust it
• The contract scanner — a daily job search that queries across LinkedIn, Indeed, ZipRecruiter and more in a single call, filters the noise, and emails you a digest
• The inbox cleaner — finds every marketing subscription buried in your mail and unsubscribes on your say-so, including on encrypted providers the commercial tools can't touch
• The learning engine — self-correction, preference memory, and pattern recognition, so it stops making the same mistake twice
• Screen vision — "Hey Jarvis, what does this error mean?" while staring at a traceback, and it reads your screen and tells you

Who this is for

You should be comfortable at a command line and installing Python packages. You do not need to be a developer, and no computer science background is assumed — every piece of code is in the book and every decision is explained.

This is not a computer science textbook. It was written by a supplier quality professional with a Ph.D. who learned Python to solve his own problem. If you've ever opened a Python file and immediately closed it again, this book is for you.

The last chapter is a roadmap, not a conclusion

What you finish with is a foundation. It runs, it works, and it's genuinely useful on day one — but it's the floor, not the ceiling. Where you take it from there is on you.

My own build went a long way past these pages: an options bot alongside the ones in the book, full Telegram control from my phone, a watchdog that restarts failed services before I notice, automated fill reconciliation on every paper trade, and a publishing pipeline that drafts blog posts and pushes them live. None of that is in the book. All of it was built on what is.

⚡ See how the voice pipeline works →

What readers are saying

Why the PDF and not the Kindle edition

This is a book you read with a terminal open beside it. The PDF keeps every code block copyable and correctly formatted — select it, paste it, run it. Kindle reflows code and converts straight quotes to curly ones, which breaks it. If you're building along, buy this one.

Prefer to read it on Kindle? Kindle edition on Amazon — $9.99 →

Format: PDF · 129 pages · 14 chapters, plus appendix and glossary
Delivery: download link emailed immediately after purchase

Build it. Own it. Run it.