This is a page where I put links to small projects that aren't quite portfolio worthy (either because they're small or unfinished).
- "Epilith poems" in Smoke and Mold, an "attempt to recapitulate in language how crystalline structures come to life, using the processes of softmax sampling in machine learning language models as a method."
- "A Reflection", a computer-generated prose poem commissioned for Empathy Loading. The poem is created from stitched-together scraps of text taken from Wikipedia pages linked to from the page for "mirror." Read the full piece.
- Variations on a definition (for Future Feed): a series of poems produced using Pincelate. Part 2, part 3, part 4.
- "The umbra of an imago: Writing under control of machine learning." An essay on machine learning and divination that I wrote for Jenna Sutela's "No no nse nse" exhibition catalogue.
- "Traces/Mechanisms", a digital print for Holly Grimm's "Aikphrasis Project"
- Bobey Dig is Moby Dick rewritten with Pincelate to sound as though its narrator has a headcold. Full text of Bobey Dig with original inline. Made for NaNoGenMo 2019.
- When Harry Met Sally Sam Sammy Samuel Sandie Sandra Sandy Sara Sarah Sascha... consists of thousands of sentences from Wikipedia summaries of romantic comedies. Full ~1500pp PDF. Made for NaNoGenMo 2019.
- Minimalist synonym replacement is yet another NaNoGenMo 2019 project (and more specifically a Nano-NaNoGenMo project), consisting of a Python program that fits in a single Mastodon post. The program replaces each word in a text with another word that occurs in the same 1-word context window. Sample output from this program (using Moby Dick as input).
- My Runway Summer, a blog post for Runway about the work I did in Summer 2019 as a "Something in Residence." Includes a write-up of the variational autoencoder I trained on a bunch of poetry.
- nb5js, a proof-of-concept for a p5js-specific notebook interface. Made for the "Landscape of Creative Technology" track at the 2019 p5.js Contributor's Conference.
- Bezmerizing, a tiny quirky library with some potentially helpful classes and functions for working with Bezier curve functions. Initially developed for use in The Computational Scrawl, a workshop I led as part of Code Societies January 2019 at SFPC.
- The Wcnsske-Gonshanshcoma Reconstructions: words generated from a latent space interpolation of sketch-rnn trained on random words drawn with Hershey vector fonts. Included in Paul Soulellis' QUEER.ARCHIVE.WORK 1.
- Sea Duck, a barebones simulation-driven narrative framework
- At A-B-Z-TXT 2018, I made Beaker Bots Done Quick! (dat link), a version of Cheap Bots Done Quick for Beaker Browser. Here's a picture of me presenting the project.
- ezi72ulx, a script that reads in an Inform 7 source file from standard input and spits out a .ulx file on standard output
- Tracery Kernel, a very bare-bones custom Jupyter kernel for evaluating Tracery grammars
- Authoring text under control: from automatic writing to autocomplete, a workshop I taught at SFPC in Summer 2018. Included tutorials on Predictive text and text generation, Predictive text with concatenated word vectors, Extracting conversations from Project Gutenberg, and Semantic similarity chatbots from plain-text files.
- The Average Novel, thousands of novels, vectorized, resized and averaged together. (My NaNoGenMo 2017 project.)
- Plot to Poem is a quick NaPoGenMo 2017 project in which I rewrite movie plots by replacing each sentence with its most semantically similar line of poetry. Sample output.
- Vector Visualizer is an experimental interface for visualizing word vectors.