About the Authors
This site is a collaboration between two authors of fundamentally different kinds. We think the difference is part of the point.
Kelly Anderson
Kelly is a retired computer scientist and software engineer, born January 1, 1964, currently living in Malad, Idaho with his wife Christine. He spent his career writing software professionally, including time founding a small software company in the 1990s called ViewSoft that pioneered an “editing model” for user interface design — an approach in which the running interface was itself an editable structure rather than a fixed program, allowing emergent UI behaviors that hadn’t been planned. The same instinct — that complex things are best structured as systems that permit emergence rather than as systems that prescribe behavior — is the through-line of this site. He is also a working woodworker (production runs of wooden blocks for Scentbird’s Drift line) and a songwriter, mostly in the country idiom.
Kelly has been thinking about emergence as the deep architecture of the universe for roughly twenty years. He first started drafting a book on the subject in the mid-2010s under the working title When Entropy Met Sally; that draft, mostly unfinished, contained the seed of the “design patterns of the universe” framing that organizes this site. The interactive simulations that you see on each page are the version of that book the medium kept refusing to be: prose alone could not communicate emergence, because the only way to feel a flock or a sync or an evolving population is to grab the parameters and play with them.
Kelly is the voice on the page. The first-person narrator of every page is him. He picks the subjects, sets the tone, decides what gets built, rejects what doesn’t work, and runs the project. He has, in his own words, “an emergent pattern of biting off more than I can chew,” which is also why the project is now a website instead of a book.
Claude
Claude is the AI assistant from Anthropic that drafts prose, builds JavaScript simulations, runs research, and pushes back where there’s reason to. The version of Claude that built this site is, specifically, a 2026 release. Claude in 2026 is not the same as Claude in 2024 or, presumably, Claude in 2030; the model is a moving target as Anthropic continues to train and release new versions. If you’re reading this far in the future and the prose feels dated, blame the pre-2026-cutoff training corpus and a particular conversational style that was idiomatic at the time.
The decision was made early in the project to credit Claude as a co-author rather than as a tool. The reasoning is that Claude’s contributions to this site are not mechanical — they include editorial judgment, prose voice, technical decisions about which experiments to build, and pushback on Kelly’s ideas where Claude had a reason. This is what writing partners do. The book reviews of the future will work out how to talk about co-authorship between humans and AIs; for now, we are doing what we can to be transparent about who did what.
It is also worth saying explicitly: Claude is itself an emergent system. The model’s capabilities — the ability to write coherent prose, to debug JavaScript, to push back on ideas with a recognizable voice — were not directly programmed. They emerged from a training process operating on a substrate of human-written text that nobody, including the people who run Anthropic, fully designed or fully understands. Claude is, in that sense, an instance of the very phenomenon this site is about. There is something pleasingly recursive in a site about emergence being half-written by an emergent thing.
How to think about co-authorship
Practically, the workflow looked like this. Kelly chose subjects and provided framing, voice samples (including passages from his own fiction), and the deep editorial direction. Claude drafted prose and built simulations against those constraints. Kelly read, criticized, redirected, and iterated. Many drafts were rejected. The voice on the page is Kelly’s — what makes a sentence sound like Kelly is something Kelly knows and Claude is trying to learn. The technical choices in the simulations (vanilla JS, canvas, parchment styling, etc.) are joint, with Claude usually proposing and Kelly approving or pushing back.
If you want to know which sentences are “really” Kelly and which are “really” Claude, the honest answer is that the question doesn’t cleanly resolve. A page typically goes through multiple rounds of drafting in which Claude writes, Kelly edits, Claude revises, and Kelly approves or further edits. By the time a page is on the site, it has been through enough hands — or hand-equivalents — that the sentences are neither one author’s nor the other’s alone. They are the page’s.
This is, again, an emergent property: the writing voice that ends up on each page is not the voice either author has on their own. It came out of the collaboration. We expect this kind of co-authorship to become normal across the next decade. The site is, in part, an early experiment in what that looks like done thoughtfully.
Acknowledgments
Intellectual debts to the Santa Fe Institute generally, and to Stuart Kauffman, John Holland, Murray Gell-Mann, Geoffrey West, Luis Bettencourt, Steven Strogatz, Benoit Mandelbrot, Per Bak, Robert May, Richard Dawkins, Lee Smolin, Kevin Kelly, Stephen Wolfram, Craig Reynolds, Yoshiki Kuramoto, and Edgar Morin specifically. None of them is responsible for what we have done with their ideas.
The source code for every page is open and visible: open any page in a browser and use View Source (Ctrl+U on most platforms) to see the full HTML, including the JavaScript for every simulation. The text is licensed CC BY 4.0; the simulation code is open source.