Short answer: no. Dreamknit patterns are not made by AI.
When you design a garment in our design tool, there's no language model guessing its way to a pattern. It's math and knitting logic that we've written ourselves, rule by rule. Nothing is random: the same yarn and the same design choices give you exactly the same pattern, every single time.
Here's why we built it this way, and how we make sure every pattern adds up.
A KNITTING PATTERN IS AN ALGORITHM 🤖
Anyone who knits knows there's a lot more to it than people think: knowledge, experience and a good dose of math. A sweater is really one big calculation.
Knitting patterns have actually always been algorithms. They've just been written in "everyday" language instead of code. Take a look:
- "Cast on 100 stitches": the number of stitches is a variable, and here its value is 100.
- "Repeat this row 15 times": repeating is a loop, and 15 is how many times it runs.
- "Size S: bind off 2 stitches. Size M: bind off 3 stitches": a condition. If size S, do this, otherwise do that.
What we do at Dreamknit is formalize the logic knitters have used for hundreds of years. We put all that knitting knowledge into a system, so you can relax, put the calculator away, follow the pattern and enjoy the craft itself 🤝
TEXTILES AND PROGRAMMING GO WAY BACK 🧵
This isn't new. In the early 1800s, Joseph Marie Jacquard developed a loom attachment controlled by punched cards. The cards decided the pattern in the fabric, and they're often counted among the very first "programs".
When Charles Babbage designed the Analytical Engine, one of the forerunners of the computer, he borrowed the punched card idea straight from the Jacquard loom. Ada Lovelace, often called the world's first programmer, put it beautifully in her notes from 1843: the engine weaves algebraic patterns just as the Jacquard loom weaves flowers and leaves.
In the same notes, she wrote something that fits surprisingly well today: the engine can't come up with anything on its own, it can only do what we know how to tell it to do. More than a hundred years later, Alan Turing called this "Lady Lovelace's objection". Our design tool is that kind of machine, on purpose: it does exactly what we've taught it, and nothing more.
And yes, the Ada Sweater is named after Ada Lovelace 🩵
WHY NOT AI? 🤔
Knitting needs exact numbers
If the pattern says "cast on 230 stitches", it has to be exactly 230 stitches. Not 228, not 232. AI models give probability-based answers. They guess what's most likely to be right, and sometimes they make up things that sound convincing but are wrong. Ask twice, and you might get two different answers.
In knitting, there's no "roughly". If you need to increase 16 stitches evenly over 100 stitches, it doesn't help to know that's one increase every 6.25 stitches. You can't knit a quarter of a stitch. The pattern has to give you instructions you can actually follow, that work out to the exact stitch and give you a beautiful result.
The knitting know-how is built in
Our algorithms don't just calculate numbers. They also make sure the instructions can actually be knitted, and that the result looks good. A lot of what's obvious to a knitter, we have to teach the computer. Raglan stitches, for example, only belong on a raglan sweater. To a computer, "raglan stitches" is just one option among many, and without the rules we've written in, it could create a sweater with set-in sleeves and three raglan stitches. That's impossible to knit 🥶
Every yarn is different
Bouclé, for example, doesn't behave like other yarns. And your gauge can vary a little. That's why our patterns have to handle small variations while the math stays completely exact. We've learned that balance through years of knitting, testing and feedback from knitters. It's not something an AI can guess.
Knitting language is its own language
The instructions in our patterns are written by us, in both Norwegian and English. Not machine translated. A garment has hundreds of instructions full of technical terms, and a translation done by AI can quickly become meaningless to a knitter.
EVERY PATTERN IS MADE FROM SCRATCH 🧮
Our patterns aren't sitting ready-made in a database. Every time you design something, the pattern is calculated from scratch, based on your yarn and your choices. The same system writes the pattern for everything from baby mittens to a men's sweater in 4XL.
That adds up to a huge number of possible patterns. As of September 2026, Dreamknit offers more than 560 million possible design combinations. Then comes the yarn: you can choose from more than 700 yarns and yarn combinations, or add your own, with 23 different gauges from 10 to 32 stitches per 10 cm. Add gauge to the mix, and we're at around 13 billion possible patterns 🚀
That would be impossible with regular, pre-written patterns.
HOW WE MAKE SURE EVERY PATTERN IS CORRECT ✅
- We knit it ourselves. New design options are knitted by us and by beta testers before they launch.
- Two sets of eyes on everything. No change to our patterns reaches you until at least two of us have looked at it.
- Automated tests. Nobody can knit 13 billion garments, so before every single change to the code, we run thousands of tests that check the math adds up.
- No silent changes. If a pattern changes, even from 230 to 232 stitches, we get notified and review the change before it reaches you.
THE PEOPLE BEHIND EVERY PATTERN 👸
We're a small team of knitters and developers in Trondheim, Norway, with one goal: to make the best knitting patterns out there. Patterns that are correct down to the stitch, easy to follow, and give you a garment that fits just the way you want.
Programming is a craft too. You can solve a problem quickly and messily, or you can learn the craft and solve hard problems in an elegant way. We've chosen the second, in both our code and our knitting.
Knitters in more than 100 countries have knitted garments from our design tool. If you have a question along the way, real people answer, and we knit too.
Every bit of feedback from you makes the next pattern a little better 🫶 Want to get to know us? Meet the team here.
FREQUENTLY ASKED QUESTIONS
Does Dreamknit use AI to make knitting patterns?
No. Our patterns are made with rule-based math and knitting logic that we've written ourselves. The same yarn and the same design choices always give you the same pattern.
What's the difference between an algorithm and AI?
When people talk about AI today, they usually mean language models like ChatGPT and other systems that have learned patterns from huge amounts of data. They give the answer they think is most likely. That's useful for lots of things, but not when 230 stitches means 230 stitches.
An algorithm like ours follows fixed rules that we've written, and gives the same answer every time. Strictly speaking, rule-based systems are also counted as an early form of AI, but that's not what most people mean by the word today.
Are the patterns written by people?
Yes. All the instructions in our patterns are written by us, in Norwegian and English. The design tool calculates the numbers for your exact yarn and your choices.
How do I know my pattern is right?
Every change to the code is tested automatically, new design options are knitted up before launch, and we follow up on all feedback. Your pattern is calculated from the gauge you enter, so knit a gauge swatch first, and your pattern will match your gauge.
Ready to make something that's yours? Try the design tool here 🩵
WANT MORE TO READ? 📚
- Ada Lovelace: The Making of a Computer Scientist by Christopher Hollings, Ursula Martin and Adrian Rice (Bodleian Library, 2018). Oxford mathematicians use Lovelace's own manuscripts to show how she learned the math behind her famous notes.
- Lovelace's own notes from 1843, for when you want to read the original.
- Maskiner som tenker by Inga Strümke (Kagge, 2023, in Norwegian). Inga is an AI researcher at NTNU, in Trondheim, and explains how artificial intelligence actually works.
