Diode font
Design Engineer · 2026
It started with the logo: I was trying out ways to set my own name, and the pixel version hooked me hardest. I adapted a single block of text, looked at it — and realized the whole site had to move over to it. One problem: a pixel font with Cyrillic that bends to every task — from a logo down to a 9-pixel caption — doesn't exist. I combed the entire Google Fonts catalog: pixel fonts with Cyrillic are rare, and every one of them is rigid. So instead of hunting for a font, it turned into an evening puzzle: draw my own. An hour was enough.
The idea
The portfolio was built around a thought: I don't just grow teams and systems — I tune them. And what gets tuned is instruments: screens, boards, panels. On an instrument, text isn't typed — it's lit. Of all the visual moves I tried, dots turned out to be the only one that rhymed with everything at once: the logo, the screen-like cards, the interactive parts.
First I honestly looked for something off the shelf. The Google Fonts catalog: 19 monospaced fonts with Cyrillic, 78 display ones, maybe three pixel fonts among them — and each does exactly one thing: sit at a single size and refuse to move. But I needed one system to carry the logo, the headings, the navigation, the counters, and nine-pixel captions. Nobody makes that.
So we draw our own. Spoiler: it took an hour.
How it works
Not a font, but an engine. Each letter is a 5×7 bitmap: rows like "01110," where a one is a lit diode. The renderer lays the text out on a dot grid and draws it with SVG circles.
Parameters instead of point sizes. The renderer has three knobs: grid step, dot radius, and bitmap upscale. At a small size the dots merge into a letter; at a large one they break apart into an honest board. That's a designer's decision, not a property of a file.
currentColor. The text is colored and animated with ordinary CSS classes — hovers, transitions, glow all work out of the box.
One language — different inputs. The same dot grid renders more than letters: pixel icons, the logo with animations (wave, equalizer, rain, scan), and 3D cubes with physics on the case cards. Right now the engine lives in 33 of the site's files.
The full set — live
So you don't have to take it on faith that this is a system and not a couple of pretty letters — here's the whole font. Cyrillic, Latin, digits, symbols. Spin the sizing: at "Caption" the dots merge into a letter, at "Board" they break apart into an honest diode screen.
[ "00000", "00000", "00000", "00000", "00000", "00000", "00000", ]
“Save to gallery” makes the glyph visible to everyone — the gallery is shared. An image settles into the grid by contrast, and the bitmap can be pasted straight into led-font.
Anyone can hide a glyph — it disappears for everyone. Anyone can bring it back too.
Nothing here yet. Draw something and hit “Save to gallery” — the glyph will show up here for everyone.
What comes out of it
The font is an engine, not a picture. The same bitmaps work as a neon sign, an airport board with a repaint wave, a counter with rolling digits, and a mask for pixel fire. Everything below is live components, not video.
An hour with the agent
The most interesting thing in this story isn't the glyphs — it's how it was made. The whole way from the first letter to a system we went through together with an AI agent in an hour: the first twenty minutes were experiments, then fittings, readability tuning, and assembling the components.
My part is the language and the calls: which matrix, where the dots should read as texture and where they should merge, what makes it into the system and what goes in the bin. The agent's part is the grind: lay Cyrillic out into bitmaps, assemble the renderer, don't break the line wraps. The hardest part turned out to be Cyrillic in 5×7 — fitting Ж, Щ, and Ю into five columns is an exercise in minimalism that we iterated on together.
It's a working model of my "design × code × AI" approach: not a presentation about the process, but an artifact of it. The best proof is right in front of you: this whole case is written in that very font.
What's next
The engine is spun out into a separate repository under MIT — take it, draw your own glyphs. A mini-builder already works: a grid where you draw your glyph with dots, save it to a browser gallery, or drop in a PNG — the engine converts the picture to dots by contrast, and from there you finish it by hand. From a font it turned into a small tool — it's higher up the page and on a separate demo page.
Why I do this
It matters to me that ideas end in something physical — a thing you can touch. The "I tune instruments" idea could have stayed a slide in a deck; instead it became the font the whole site is written in. Puzzles like this keep my hand on the technology and put my taste to the test: an hour of work, and it's dead easy to tell a systemic solution from decoration by the result.