LLM Party

I built small, silly experiments. Each one taught me something a tutorial wouldn't: where the model is brilliant, where it lies with total confidence, where it forgets, and where it needs one plain line of code to keep it honest.

Playing is the fastest way to learn what this technology can and can't do. And you should learn it, whether you want to use it, criticise it or refuse it.

Working with AI is as much about design as it is about technology. The experiments that work best aren't the most complex: they're the ones that feel natural.

Bigger models aren't always better. Many of these work fine with small, local models. Match the model to the job.

Next.js, React and TypeScript, with models from several providers (Anthropic, OpenAI, Google, Mistral, Groq and local ones) behind one API layer. Framer Motion for the animations, and a homemade retro design system on Tailwind.

Curious, have feedback, or want to build something together? Write to llmparty@proton.me.

Models are fairly expensive. If you want to help with the cost, I accept tips on Ko-Fi ☕.

Backgrounds from free-game-assets and their city set. Icons from pixelarticons. Font: Pixelify Sans.

Music: Comigo, nomiqbomi, josefpres, joshuuu.

Sounds: zipdisq™, tom_woysky, GiocoSound, mickthemicguy, joao_janz, littlerobotsoundfactory, taranp, legnalegna55, elmasmalo1, theTone.

Racing pack: Kenney.