Building BookAI: An AI-Powered Memoir Writing Platform

Some stories need to be told. Not for fame or fortune, but because they matter. Because someone survived something, witnessed something, or uncovered something that others need to hear.

This is why I'm building BookAI.

Why a Memoir?

I've lived a life that doesn't fit neatly into boxes. Born addicted to heroin at the University of Michigan to a mother in prison. Adopted into a family where I had to push my 350-pound father down the stairs at age 8 just to survive. Became a federal whistleblower, stood up for the law, and waited nearly four years for my day in court - only to be pressured into settling. Then forced into bankruptcy through no fault of my own when a PPP loan was approved but mysteriously never arrived, while the person closest to me seemed to be in on it.

For years, I told myself I'd write it all down "someday." But someday never comes when you're busy trying to survive, rebuild, and figure out who you can actually trust. The memories don't wait - they fade, blur, and sometimes disappear entirely.

I needed a system. Not just a blank document staring back at me, but a tool that could:

1. Capture events as I remember them - Quick entries without pressure to write prose
2. Organize the chaos - 45+ life events spanning decades
3. Store evidence - Documents, photos, records that corroborate the story
4. Help me write - AI assistance to turn bullet points into narrative
5. Protect the truth - Secure storage with sharing controls for lawyers when needed

The Tech Stack

BookAI runs entirely on Cloudflare's edge infrastructure. The entire platform costs less than a cup of coffee per month to operate.

Frontend: Astro

Astro was the obvious choice. It's fast, flexible, and lets me mix static pages with dynamic islands of interactivity. The timeline view, chapter editor, and evidence gallery all benefit from Astro's partial hydration approach.

Database: Cloudflare D1

All the data lives in D1, Cloudflare's SQLite-at-the-edge database:

• Events - Raw memories with dates, descriptions, and categories
• Chapters - The structured narrative, 10 chapters planned
• Evidence - Metadata for documents and photos
• People - Cast of characters with their roles in the story
• Writing Sessions - AI interaction history for continuity

Storage: Cloudflare R2

Evidence files - scanned documents, photos, court records - live in R2. Private by default, with secure sharing links for lawyers and trusted readers.

AI: Cloudflare Workers AI

The writing assistant uses Llama 3.1 70B through Workers AI. It helps with:

• Expanding bullet-point events into narrative prose
• Maintaining consistent voice across chapters
• Suggesting connections between events
• Drafting difficult scenes with sensitivity

The AI doesn't write the memoir - I do. But it helps me get unstuck, find the right words, and keep momentum when facing memories I'd rather not revisit.

The Workflow

1. Event Capture

When a memory surfaces, I capture it immediately. Just the basics:

• What happened
• When (approximate is fine)
• Who was involved
• Category (family, trauma, triumph, legal, career)

No pressure to write beautifully. Just get it down.

2. Evidence Linking

For events with documentation, I upload the evidence:

• Court records
• FOIA responses
• Photos
• Communications

Each piece links to its relevant events, building a corroborated timeline that can't be gaslit away.

3. Chapter Planning

Events get assigned to chapters. The 10-chapter structure emerged naturally from the life events themselves.

4. Writing Sessions

When I'm ready to write, I pick a chapter, review its linked events, and start a writing session. The AI helps expand events into scenes, but every word gets my review and revision.

What I've Learned

1. Structure enables creativity - Having a system for capturing events removed the pressure to "write well" immediately
2. Evidence matters - Linking documents to events builds confidence that the story is accurate and can't be dismissed
3. AI is a tool, not a replacement - It helps with the mechanics of writing, but the voice and truth must be mine
4. Edge computing democratizes infrastructure - This entire platform runs for pennies, making personal projects viable
5. Building beats waiting - Instead of waiting for someone to give me a platform, I built my own

What's Next

The events are captured. The chapters are planned. Now comes the actual writing.

Sometimes the best projects are the ones we build for ourselves. The skills transfer, the lessons stick, and occasionally, the result matters more than any client work ever could.