Small research pages built the fast way: pull the data, read all of it, then ship one self-contained page. Culture, design and quiz corpora, mostly. Each one is a finished argument, not a dashboard.
The full corpus of a Russian UX-research channel — 107 posts, 291k characters — scraped and measured to reverse-engineer one author’s voice: the seven-step post skeleton, seven hook types, the inventory of theorists she leans on, emoji as list markers, and the anti-patterns that give a fake away.
What Kyle Chayka actually claims about the AI look taking over the web: the six tells, the designers he quotes, the vendor’s own admission. Set in that very style, as its own exhibit.
Русская версия разбора: что именно Кайл Чайка утверждает про «ИИ-эстетику», какие шесть признаков он перечисляет и кого цитирует.
Open source visual schemes to swap a project into — neo-brutalism, pixel, glass, retro-OS, terminal, vaporwave — each with a live specimen, its stack and whether it is still maintained.
Do readers like the questions they can answer? A 36-question package scored two ways — tournament solve rates against Telegram emoji reactions. Spoiler: the two barely correlate.
Vocabulary richness of quiz-question authors. 6,701 packs, 436k questions, 149 reliable authors, ranked in one interactive infographic.
How rich is the vocabulary of Russian stand-up comics? 321 concerts scraped and lemmatised, measured on a 25k-lemma metric, ranked head to head.
Working notes also live here, unpolished: notes on AI design aesthetics (markdown).