Artificial Intelligence

I remember a time when AI was a dirty word. Now it's a dirty word with a lot of money behind it. Some of the stuff here is from when AI was still a dirty word, some is from later. Some is from in between when LLMs started to work but were still quite bad.
Deep Learning Cookbook

Deep Learning Cookbook

I wrote a book for O'Reilly about Machine Learning. It will teach you a wide range of techniques, using projects similar to the sort of thing that you can find on this website. Word2Vec, GANs, music analysis. You should buy it! Oh and all the code is on GitHub.

The Web of Babel

The Web of Babel

A browser for an internet that doesn't exist, inspired by Borges' Library of Babel. Type any URL and an AI hallucinates a plausible web page. Click links to keep browsing.

The world this wiki

The world this wiki

The idea of LLM Wiki applied to a year of the Economist. Have an LLM keep a wiki up-to-date about companies, people & countries while reading through all articles of the economist from Q2 2025 until Q2 2026.

The AI not taken

The AI not taken

Based on OpenAIs DALL-E-2, The AI not taken generates illustrations for well known poems. For each poem there is an associated artist (usually from the same time, maybe same style) that is used to create an image in that style matching each line of the poem.

Processing Evolved

Processing Evolved

Start from a simple square and follow different evolutionary paths to create a complex animation. Each step the AI will create a new animation based on the previous one and your prompt.

Torreable

Torreable

An MCP server that lets AI agents like Goose publish static websites as Tor hidden services (.onion sites)

World Pop

World Pop

Inspired by Gap Minder, this shows the population of the countries of the world by morphing their shapes from 1880 to 2020.

Gram Zoom

Gram Zoom

Gramzoom uses the key element of the Style Tranfer Algorithm to create an infinitely zoomable movie from any image. It works best with images that are self similar, but anything will really do. Can cats look evil? It might just break the Internet.

Universal Numbers

Universal Numbers

Create universal numbers by comparing the edit distance between all numbers from Wikitravel's phrasebooks and for each picking the 'median' one. As a side effect, create a tree an evolutionary tree.

Football Predictions

Football Predictions

Monte Carlo tournament simulator. Reads pre-tournament ELO ratings scraped from eloratings.net and runs 10,000 simulations of the groups and bracket. Edit any team's rating or fix match scores to see how the probabilities shift — handy for cheering on your favourite even when the model disagrees.

State animals imagined by AI

State animals imagined by AI

Generated using stable diffusion, here we have the states and DC each represented by an image of their state animals. Another historical example of how revolutionary this was at the time and now trivial for anybody with access to their favorite chatbot.

Map of

Map of

Color a world or US map by how semantically close a word is to each country (or state) name. Built on Google News Word2Vec. The geography of meaning ends up remarkably readable: "vodka" lights up Eastern Europe, "desert" the dry bits of the planet.

Styled Museums

Styled Museums

Using the Artistic Style algorithm to restyle pictures of museums in the style of their most famous painting. This way you get immediately an idea of what to expect at these museums. Uses data from Wikipedia, Wikidata andthe Wikistats. Uses Anish Athalye Tensorflow implementation

Deep Ink

Deep Ink

An implementation of deep dreaming where the network is restricted to just black and white and starts with a blob in the middle. It forces it to draw from the center and create ink like patterns.

Mandylion

Mandylion

Using Recurrent Neural Networks to generate Icons and Hieroglyphs. The icons are encoded in a machine learnable format.

Marconi

Marconi

An implementation of a Spotify-like song radio based on Word2Vec. Based on a large set of crawled playlists and using those playlists as sentence equivalents. The Word2Vec algorithm then produces a vector per song.

Twitter Bot

Twitter Bot

This project uses the twitter api to automatically generate tweets based on what is commonly tweeted. It listens to the general tweetstream and captures fragments. By randomly recombining those fragments something that reads almost like real tweets appears. It also gives an interesting insight into what the average user uses twitter for.

For fast-acting relief, try slowing down.