The answer was messier, nerdier, and way more interesting than I expected. If you care about Japanese lettering, logo balance, and why kanji feel so alive on the page, this one’s for you.
Know that feeling when a logo looks almost right, but one stroke throws off the whole mood? That was the starting point here. The brief was simple: create a symbol for a Japanese learning app, use Claude as a design partner, and see whether an invented kanji could feel believable enough to live beside real characters.
I still remember exactly how quickly the experiment became less about “AI art” and more about type. The goal wasn’t to make something cute. It was to make something that respected the visual logic of kanji, while still being clearly fictional. That tension is what made the process so nerdy, and so fun.
Claude, a Japanese learning app, and the idea of an invented kanji
Why even invent a kanji in the first place? For a Japanese learning app, the answer is branding with intent. A symbol built from kanji-like structure can hint at language study, reading, discipline, and precision, without falling back on the usual schoolbook imagery.
A great example: many Japanese apps use clean wordmarks or mascots, but a kanji-inspired logo can create instant atmosphere. It says, “This app is about script, not just vocabulary.” In other words, the logo itself becomes part of the learning story.
Claude was the co-designer in this case, not the final decision-maker. That matters. The model could suggest forms, explain stroke tension, and compare different balances, but it couldn’t feel whether a character looked too stiff or too ornamental. That’s where the human eye still wins. Personally, that’s the most interesting part.
For learners, this also opens a small but useful window into Japanese visual culture. Kanji aren’t just meanings on paper. They have weight, rhythm, and spacing, or ma, the same kind of quiet breathing room you notice in a good ryokan garden or a poster by a great Tokyo designer. My tip: if you study kanji, look at their silhouettes first, not only their dictionary meanings.
Five design rounds with Claude: from rough sketch to believable form
By the way, the real work happened across five design rounds, and each one changed the direction in a different way. The first pass was all about shape. Claude suggested overly literal “kanji-ish” forms, which looked clever for a second and then collapsed under closer inspection. Too many strokes, too much noise, not enough structure.
Take the second round, for instance. The prompt shifted towards simplicity, and the character got cleaner, but now it felt generic. It could have been a stamp, a classroom icon, or a stock symbol. The problem wasn’t detail, it was personality. Real kanji often look controlled, but never dead.
While we’re at it, round three was where spacing started to matter. Claude began proposing how thick a main stroke should be, how much negative space could sit under a radical, and where the eye should rest. That’s classic Japanese typography thinking. Even when a symbol is fictional, the empty space has to feel intentional. No wonder, because Japanese design often treats silence as part of the layout.
Round four got more experimental. Claude tried asymmetry, a slight slant, and a more calligraphic rhythm. Some versions looked strong, but they started drifting too close to handwritten brush art. The issue there is trust. If the line gets too painterly, the logo stops feeling like a stable sign and starts feeling like an illustration. Wild, isn't it?
Then came round five, the one that finally clicked. The form was pared back again, but not flattened. It had a clear vertical axis, a memorable top-to-bottom flow, and enough internal logic that it felt like it could exist in a Japanese font family. Not bad, right? From my experience, that’s usually the sweet spot: less invention, more discipline.
What changed between rounds
Fun fact: the biggest improvements weren’t visual flourishes, they were structural corrections. Claude helped test stroke weight, read spacing, and compare variants the way a designer might compare font proofs. The difference between “looks Japanese” and “could pass as a real sign” is tiny, and five rounds were just enough to expose that gap.
Seriously though, this is where an AI can be useful for typography brainstorming. It can iterate quickly, react to constraints, and keep track of subtle changes across versions. But it still needs a human who knows when a stroke is doing too much. Personally...
What an invented kanji teaches Japanese learners
Unlike what we’re used to, invented kanji can be a surprisingly good study exercise. They force you to think about radicals, balance, and visual hierarchy instead of memorising readings only. If you know the basics of bushu, you start noticing how much of kanji recognition is pattern memory.
For example, a learner might spot a left-right structure, a top-bottom structure, or a box-like enclosure and immediately guess function before meaning. That’s not a trick, it’s how real reading works in practice. Japanese texts are full of shorthand visual cues, and trained eyes learn to scan them fast.
What I especially like: this kind of logo experiment makes the gap between “writing” and “design” obvious. A kanji can be perfectly legible and still feel wrong in a logo. Or it can be invented, but so well balanced that the eye accepts it for a moment. That tension says a lot about why Japanese type design is such a deep field.
For Japan fans, that means there’s more here than a playful branding test. It’s also a reminder that Japanese script lives in a wider design culture, from packaging and shop signs to game logos and app icons. If you’ve ever stared at a sign in Shibuya and thought, “That spacing feels good,” you already know the kind of nerdy pleasure this experiment is chasing. From my experience...
My verdict: Claude didn’t magically create a perfect kanji, and that’s exactly why the process was worth watching. The five rounds showed how much refinement a good symbol needs, even when the symbol doesn’t exist in any dictionary. That’s the real lesson for learners and designers alike: in Japanese lettering, structure is meaning.
Frequently Asked Questions
What was the goal of the logo experiment?
The goal was to create an invented kanji for a Japanese learning app and see how far Claude could help refine it across five design rounds.
Why use an invented kanji instead of a normal logo?
An invented kanji can suggest language, structure, and culture in a single symbol, which fits a learning app very well.
What did the five design rounds improve?
They helped clarify stroke balance, spacing, personality, and overall legibility, until the symbol felt closer to real Japanese typography.
