Infinite Wiki ⁕ Every word is a hyperlink. Every description is generated in real-time (in ~1 second) ⁕ Runs on Gemini 2.5 Flash Lite. ASCII diagrams using 2.5 Flash

Here's another run. Works sweetly on mobile too. Try here

You can search for any words or phrases (more in the thread below); or jump to a random word

Reaaally enjoyed some of the ASCII output that came through. Here's some highlights. The prompt was to create an ASCII visualization where the visual form embody the word's essence

Aesthetic and vibe deeply inspired by friend Sharon Park 🩵🩵🩵

pretty much a dictionary then haha. nice ascii art for each word!

@beanosthekid aha yes but it can also do proper nouns, phrases and sentences if you search for em

man super sweet @Kimi_Moonshot Kimi K2 made this in 30s: (don't look too closely - made at

HOLYYYYYY 😍😍😍😍 WHAT??? SO GOOD

@tinaz0ne 🥰🥰🩵🩵

This needs mouse text highlight selection to create a new link rather than clicking individual words.

Prior art but love the aesthetics :)

jesus christ i had this idea yesterday.

ahaha synchronicity and the universal consciousness is very real

welcome former lurker, banger

🫶 i just recently watched an old talk of yours that inspired infinite backrooms that ended up inspiring this other project i did

utterly epic, i'll have to peep that. would love to mess around sometime! i wanna see what dat base model do

I love this idk what to do with it, but I love it

I've been experimenting with this too with my product, Emergent Mind, which is focused on helping people discover and learn about AI/ML research on arXiv. For example, if a user pulls up a paper, it generates a summary that includes linked technical terms that users can click to learn about them. One learning I had along the way is that it's a better user experience not to always generate the content on the fly; cache previously generated responses and show them immediately to future users, that way users don't have to wait for the content to generate. Also, one thing to try is to generate the response, then prompt the LLM again to identify encyclopedia-worthy terms in the content, then just linkify those terms, that way you're not linking common words like "these", "of", etc. Here's an example on my site if you want to check it out: Great work though, I love seeing creative LLM UI experiments like this 👏.

Great concept - I think the ascii is redundant and degrades ux (text is pushed down while you're reading it, and it just takes longer overall)

And the model is trained on actual wiki's. So its a wikipedia wrapper? 🫣

fun!

Dear @dev_valladares, What you have showed here is amazing. I truely congratulate you from my heart. {Art + Science} => Divine We have only had Tech separately thus far. Bravo. 🫶🙏

Legendary.

Awesome! Reminds me of @Aizkmusic WikiTok a bit. We need more of these.

This is actually pretty nice

Reminds me of HyperCard, but live. Genius! (And the ASCII Art is wonderful). "Described by creator Bill Atkinson as a "software erector set",[6] HyperCard is based on the concept of a "stack" of virtual "cards".[7] ...Users browse the stack by navigating from card to card, using built-in navigation features, a powerful search mechanism, or through user-created scripts.[8]"

That’s a really neat idea and a great showcase on Gemini speed.

Thanks to you I made best language learning tool with switch translating text to PL DE and with selector of two levels of text (normal and 9 year old kid). So fun!

ha! amazing, great to hear :)

i love this so much... for some reason weights heavier on art than ai... will allocate some time to play with it as it looks like a DEEP rabbit whole "thingi magigi" 😅

Beautiful!

@shaobuilds

This is so cool @dev_valladares !

In 30 months, instead of the ASCII, you can just have Veo 9 generate your favorite documentary narrator to present the topic while Imagen 10 is renders a full-on photo or artwork of the concept while Gemini 8.5 Flash Lite spits out a full book on the topic

This is simply amazing

this is incredible

whats the system prompt for the ascii art?

const artPromptPart = `1. "art": meta ASCII visualization of the word "${topic}": - Palette: │─┌┐└┘├┤┬┴┼►◄▲▼○●◐◑░▒▓█▀▄■□▪▫★⟨⟩/\\_| - Shape mirrors concept - make the visual form embody the word's essence - Examples: * "explosion" → radiating lines from center * "hierarchy" → pyramid structure * "flow" → curved directional lines - Return as single string with \n for line breaks`;

cool! thanks

That's cool actually

This must be crazy for documentation and journaling.

<33333

big fan of you and your work with QRI!!!

thanks :D

What's the use case?

what an abomination of 400KY of human knowledge and collaboration, but also very cool.

people keep reinventing delve