LLMs: Quilting with Words
Many of you are aware that I am a technology enthusiast with a particular focus on artificial intelligence and its impact on reshaping our world of work. But what you may not know is that I am also a traditionalist and a creator of quilts. Yes, I buy yards of fabric, cut it into small shapes, and reassemble those pieces into something beautiful and functional. Quilting is a tactile, creative process that allows me to slow down and reflect, especially as I am also deep in the final stages of writing my dissertation on AI.
Over the past few weekends, I found a few hours to relax with my hobby. And, true to form, while sewing, my thoughts drifted to LLMs. The more I thought about it, the more I realized how similar these two seemingly different pursuits really are.
Follow along with me.
Quilters buy fabric by the yard, sometimes by the fat quarter (an 18x22-inch cut), and sometimes as five-yard bolts. Over the years, I have amassed a comprehensive library of fabric, meticulously organized into shelves, bins, and totes by color, texture, and print. Each piece of fabric has its own unique qualities of tone, texture, pattern, and vibrancy, making it distinct, much like individual works of writing or speech.
When we quilt, we slice and dice this fabric into smaller pieces, such as squares, triangles, strips, and hexagons. We don’t preserve the fabric whole; instead, we deconstruct it and then reassemble it, with intention, into something new, like a quilt top that no one has seen before. Often, we do this by following a pattern, but even with the same pattern, no two quilts are ever identical. The individual fabric pieces may be familiar to many quilters, but the combination, placement, interpretation, and quilting are uniquely our own.
This is strikingly similar to what large language models (LLMs) do with language. An LLM is trained on vast libraries of text from literature, journalism, social media, and academic papers, each with its own voice, tone, and style. When prompted, an LLM slices through this massive corpus of language and reassembles words and phrases to generate something new. Like quilting, the process is not about copying and pasting from the original sources but about remixing, reshaping, and creatively assembling based on patterns it has learned.
Now, here is the fascinating parallel: quilters around the world often buy the same fabrics. When a new fabric line is released, many of us purchase a coordinated bundle of 12 to 24 fabrics. Thousands of quilters may be working from the same raw materials, but no two finished quilts look exactly alike. The choices we make, from which fabrics to emphasize, which patterns to use, and how to arrange colors and shapes, reflect our own taste, creativity, and experience.
The same is true of LLMs. All LLMs draw from similar sources of language. We all prompt them in the same base language. Yet the outputs vary dramatically based on how we interact with them. Your tone, your phrasing, and your intent all influence the model’s response. Even when two people use the same model with the same knowledge base, they rarely produce identical outputs because we, as human users, shape the conversation.
And here is where psychology offers an additional insight. Researchers have long noted what is called the contagion of ownership or endowment effect: when we interact with something, be it a physical object or an idea, we inevitably leave a trace of ourselves on it, and in doing so, it becomes more meaningful to us (Kahneman, Knetsch, & Thaler, 1990). In quilting, even when using a commercial pattern, we often modify it by choosing different fabrics, adding borders, or adjusting sizes, as we want the quilt to reflect our personal style. The same cognitive drive is at work when we prompt LLMs: we unconsciously shape the output to reflect our voice and intent, making the result feel uniquely ours.
In both cases, quilting and prompting, we cannot help but leave a signature on the final product. Our aesthetic choices, our biases, and our creativity all become embedded in what we create. And that is a beautiful thing.
It serves as a reminder that even in an era of increasingly sophisticated AI, human creativity, interpretation, and individuality remain central. The machine may help us assemble the pieces faster or explore new combinations, but the artistry is still very much ours.
As I sew another quilt or write another dissertation paragraph, I am reminded that creation, whether with fabric or language, is always a dance between the raw materials and the human mind. And every stitch, every word, carries a small amount of us forward into the world.