Twenty-seven tiny worlds, each one a different AI trick.

Describe a thing; the model renders it as ASCII art. Side-by-side gallery, copy-to-clipboard, gallery export.

Tee up a problem; the model rewrites your snippet in as few characters as possible. Stroke counter, par rating.

Type a premise, get a four-panel comic script with captions, panel directions, and punchline polish. Ink on newsprint vibe.

Pick a motion; two models argue opposing sides turn-by-turn. Oxford-debate structure, rebuttal, and a closing verdict.

Log a dream; three distinct analyst lenses (Freudian, Jungian, somatic) interpret it side by side. Mystic-horoscope styling.

Feed in N emojis; get a story where each emoji is a chapter heading and appears literally in its section. Storybook UI.

Choose-your-own-adventure with Ollama as Dungeon Master. Branching narrative, inventory, dice rolls, save state.

One question, five personalities (sceptic, optimist, artist, hacker, grandma) — each gives their take in their voice.

Pick a topic; get Cantonese rap bars with rhyme markers, flow notes, and a beat-grid timing view. 押韻 · FLOW · 節拍.

Snap a photo or list what's in your fridge; the model invents a recipe with steps, timings, and an ingredient balance chart.

Two AI comedians. One topic. Three rounds. Pure savagery — plus a real refactor suggestion that nobody asked for.

A creative-studio brainstorm: 10 slogan drafts, scored on clarity, originality and punch, then re-ranked by a critic model.

A 4-stage AI murder mystery: write the case file (5 suspects, 3 clue tiers, hidden solution), then interrogate any suspect, then deliver a verdict. Multi-turn.

Type your current mood and pick a genre. The model drafts an 8-track playlist with title, artist, year, and a one-line "why it fits". Gradient tiles, shuffle, copy.

Pitch a 1-sentence idea. The model delivers a brutal VC-style roast and a constructive counter-take, side by side. Steelman mode defends your idea on demand.

Pick a theme and difficulty. The model drafts a 5×5 (easy/medium) or 7×7 (hard) crossword with numbered cells, across + down clues, and reveals. Type to play.

Type any boring object. Get three escalating pitches: deadpan PSA, unhinged infomercial host, and whispered ASMR luxury. Plus a Pitch-of-the-Day rotation.

Type one sentence. Five parallel streams translate it into pirate, valley girl, Shakespearean, southern belle, and robotic English. Same meaning, wildly different vibes.

Type a prompt. Get five six-word stories, each self-rated for gut-punch, sorted by score, winner gilded in gold. Inspired by the legend, weaponised by AI.

Type any object. Get sommelier-grade tasting notes: nose, palate, finish, pairings, cellar aging potential, vintage. Six pre-loaded absurd objects in the Cellar.

Paste your resume. The model delivers a one-line verdict plus five color-coded sections: The Hook, Buzzword Triage, Achievement Inflation, Chronology Story, The Hired Reading. Three tones from brutal to kind.

Paste any contract, ToS, EULA or lease clause. Get a one-line summary, a 1-10 risk badge, red flags, fair-enough clauses, things you can negotiate, and five questions to ask before signing. Four pre-loaded samples.

A 16-team single-elimination trivia bracket. AI generates every match's question in real time, eight bot personas race you, commentary after each round. One player, vs-bots, or pure watch mode.

Type a city; the page builds a tiny 3D world — sky dome, drifting clouds, sun, rain — to match the live forecast. three.js + free GLB models.

Pick a theme; the AI drafts 14-18 entries with clues, then a Python solver builds an 11×11 grid that actually interlocks. Three sample puzzles.

Topic + audience → 3 title options, big idea, 5-section structure with timing, 3 stories, mid-talk twist, CTA, 3 hard Q&A pre-empts, and a 1500-word speaker script. Two samples.

Topic + guest + format → cold open, intro, 8-10 question interview, pushback, mid-roll ad read, listener questions, outro, show notes, and a 60-second social clip. Three samples.