Innhold Show
A phone. An AI subscription. And a box of 133-year-old poems. That was enough to take me from no publishing experience to a book available through Norwegian bookstores – with less than a week of actual work.
In part 1, I told the story of finding the poems written by my great-great-great-grandfather, Gustav Johan Carlén. My sister Lillian found 33 handwritten pages from the 1890s in a dusty box in the garage.
This is how those pages became a book. I had no publishing contacts, no experience of publishing a book and no desire to spend a fortune.
Where I started
I had no idea where to begin. I had 33 poems – some handwritten, others typed by my grandfather in the 1950s or 1960s. I knew nobody in publishing. I had never published a book before.
What I did have was a phone with a camera, a Claude subscription and the slightly ridiculous belief that this had to be possible.
It was easier than I expected.
Step 1: From handwriting to digital text
The first challenge was the handwriting. Gustav Johan had written with pen and ink more than a century earlier. The letters were narrow and slanted, and the ink had faded. Even I – his own descendant – struggled to read all of it.
I photographed every poem with my phone. That alone took a few hours because I needed good lighting, the right angle and an image clear enough to read.
Then I gave the images to Claude. The AI transcribed the handwriting into digital text. It was not perfect, and I had to correct it at times, but it did about 90% of the work in minutes rather than hours. I copied the text into Google Docs, and suddenly I had all 33 poems in a format I could work with.
Think about that. A century of yellowed paper and faded ink, digitized in an afternoon.
Step 2: Understanding the poet
This was where something unexpected happened. I used Claude Code to analyze the poems, not only linguistically but thematically. Who was Gustav Johan? What did his poems reveal about his life?
The AI helped me see patterns I had missed: recurring themes, biographical clues hidden in the verses, and references to places and events.
From the poet’s own words, I reconstructed a biography with AI as a research partner. It was not complete – these were poems, not a diary – but it was enough to understand more about the man behind the words: his childhood as a shepherd boy, his years as a factory worker and his struggle for dignity in a society that did not recognize his people.
I also found historical context about the Romani/Traveller people in Norway at the time: the discrimination they endured and the laws used to persecute them. That gave the poems a depth I had not seen before.
Step 3: Structure and spelling
Gustav Johan wrote in the Norwegian of the 1890s. Some words were archaic. Others looked misspelled – but were they? Perhaps that was how people wrote at the time. Perhaps it was dialect.
I had choices to make. Should I modernize the language, correct the “mistakes” or preserve the original?
I chose a middle path. Claude helped me identify language that belonged to the period and distinguish it from obvious writing errors. I kept most of the text as it was because the poet’s voice mattered more than grammatical perfection. Where the meaning became unclear, I added careful footnotes.
Then came the structure. How do you organize 33 poems into a coherent book? Claude suggested thematic sections: childhood, travelling, social criticism and hope. I wrote a foreword and connecting passages to give readers context, with AI as my sparring partner throughout.
Step 4: The cover that surprised me
This was the part that genuinely made me stop and stare.
I needed a front cover and a back cover. Normally, that would have required a graphic designer, Adobe skills or at least a few days of tutorials.
Instead, I used Nano Banana Pro, an AI image generator. I described what I wanted: a travelling man, a forest and an image that captured something of Gustav Johan’s life.
The first attempt was not perfect. Neither was the second. After a few iterations, however, I had a cover I was proud of.
The back cover impressed me even more. I gave the AI the blurb that Claude had helped me write and asked for a complete back-cover design with the text integrated. The result was a finished, print-ready image. No Photoshop. No design experience. Just a written description and a few attempts.
Step 5: Preparing the book for print
Margins. Fonts. Page layout. Type area. Words I barely understood before I started.
Claude guided me through all of it: the margins printers expect, the fonts that work for poetry, and how to make facing odd- and even-numbered pages mirror each other correctly. Technical details that would have taken me weeks to learn on my own were dealt with in hours.
The result was a PDF ready for print-on-demand.
Step 6: From a file to bookstores
The final piece was figuring out how to get a book into the world.
The AI recommended BoD.no, or Books on Demand, a Norwegian print-on-demand company. That meant no inventory costs and no minimum print run. A copy would be printed whenever someone ordered one.
BoD arranged the ISBN and distributed the book to Norwegian booksellers including Norli, Ark and Adlibris. Suddenly, Gustav Johan’s poems were available across Norway.
At the time, I paid NOK 1,000 for distribution to Norwegian bookstores. That was the price for my project in Norway, not an international price quote.
What it cost
Here is the actual cost breakdown:
- AI services: A Claude subscription I already had
- Image generation: A few dollars with Nano Banana Pro
- BoD distribution: NOK 1,000
The total was under NOK 2,000, from the first idea to a book available through bookstores.
And the time? The actual work took less than a week, spread over a longer period – a little here and there between other things. Most of the time went into photography and quality control, not production itself.
Why this matters
I am not telling this story to show how clever I am. Quite the opposite. I was not particularly skilled at any of this. I had no book-publishing experience, no design expertise and no contacts in the industry.
I had tools that lowered the barrier dramatically.
AI is not only for writing marketing copy or code. It can preserve material that might otherwise disappear.
How many boxes like this are sitting in Norwegian attics right now? Letters from a grandmother. A great-grandfather’s diaries. Photographs filled with unidentified faces and stories nobody remembers. Poems that were never read outside the family.
A few years ago, doing something serious with that material required specialist knowledge or expensive services. Now one person can do it in a week with a phone and an AI subscription.
AI as a tool for cultural preservation
This is not the first use people tend to imagine when they hear “artificial intelligence.” Look at what AI actually helped me do:
- Transcribe old, difficult handwriting
- Understand historical context and language
- Structure and edit text
- Create visual material
- Navigate technical production processes
- Find a practical publishing and distribution route
Taken together, those jobs once demanded several specialists and a great deal of time. I still had to make decisions, correct mistakes and check the work, but the threshold had changed completely. What I needed most was curiosity and the willingness to try.
The result
The book is called Lys på veien – en reisende fants etterlatte dikt. The title can be rendered in English as Light on the Road – The Posthumous Poems of a Romani Traveller. Gustav Johan Carlén is the author. My sister, Lillian Bauge Carlén, and I published it.
You can buy it from Norli, Ark, Adlibris and most Norwegian bookstores. Six months earlier, it had been a collection of papers in a dusty box in a garage.
Do you have a box in the attic? An old letter from your grandmother? Forgotten photographs filled with unknown faces?
Perhaps somebody in that box is still waiting to be heard. AI can help you find them.
More than a century ago, Gustav Johan wrote: “Det lyse maa før andre, som uti tusental foruten lys nu vandre i livets skyggedal.” Roughly translated: “The light must go before those who, in their thousands and without light, now wander through life’s valley of shadows.”
He hoped his words would bring light to others. With a little help from modern technology, they finally do.
The tools I used
- Claude and Claude Code: Transcription, analysis, writing and formatting
- Nano Banana Pro: Front- and back-cover design
- Google Docs: Word processing
- BoD.no: Print-on-demand and distribution
Part 1 of this story introduces Gustav Johan Carlén – the Romani Traveller, shepherd boy, factory worker and poet who was my great-great-great-grandfather.
Frequently asked questions
Can AI transcribe old handwritten poems?
Yes, if the photographs are clear enough, but the result still needs human checking. Claude did about 90% of my transcription work, while I corrected errors and compared the digital text with the 33 original pages.
How did I preserve the poet’s original language?
I kept most of Gustav Johan’s 1890s Norwegian unchanged. AI helped me distinguish period language and dialect from obvious mistakes, while I used careful footnotes where the original meaning might otherwise be unclear.
How did the book reach Norwegian bookstores?
I used BoD.no for print-on-demand, ISBN administration and distribution. Copies are printed when ordered, so I did not need to finance inventory or commit to a minimum print run.
What should I do with old family papers in the attic?
Start by photographing or scanning everything clearly, then transcribe and verify it before changing the language or structure. Keep the originals, document your editorial choices and treat AI as a capable assistant rather than an unquestionable authority.
