Only variety can absorb variety. Ask what in our tradition is proved, and that sentence comes back, usually with Ashby’s name under it. Ashby wrote something else: “only variety in R can force down the variety due to D; variety can destroy variety” (1956, §11/7). The sentence we quote is Beer’s.
Chapter 3 of Punch and Judy, “The Law We Quote Is Not the Law Ashby Wrote”, is about that gap. On 28 September it got something no chapter on this site has had before: a cheat sheet beside it, one page of context and four pages that go deeper. The Introduction to the book went up the same day, as a chapter you can read here.
The cheat sheet is the smaller of the two. We think it matters more.
What we mean by a theme
A theme is a claim or a question we open to the public, usually one chapter. Chapter 3’s theme is the piece of our inheritance we are most confident in: the law of requisite variety, Beer’s compression of it, and the good regulator theorem we lean on almost as hard.
Until now, opening a theme meant publishing the chapter. That assumes a reader who knows this literature already, or one who will work through a draft that cites page numbers of books they may not own. Both kinds of reader exist. Neither should have to start there.
The plan is that each theme we open comes with three layers of context underneath it.
An introduction to the theme. The cheat sheet overview for Chapter 3 does this in about a page: which area the chapter sits in, what it separates, what it means for the model and for the book, and where to start reading. We wrote it for someone who has never met Ashby, and for someone who has taught him for twenty years and wants the short version. If it fails one of those two readers, tell us which.
The neighboring themes. Most of what a chapter needs lives next door. Chapter 3 needs information theory, control theory and outcomes from 1970 to 2025, and a reader should not have to find those from scratch. We have mapped eight areas around the program, each with its own cheat sheet: cybernetics and the Viable System Model, philosophy of science and metascience, general systems theory and rival minimal sets, organization theory, design and failure, cognitive science and philosophy of mind, neuroscience, network science and systems biology, complexity science, ecology and evolution.
Deep dives, eventually. One page per contested object, with the sources, the page numbers and what each source says. Four are up for Chapter 3: Ashby’s law as written, the good regulator theorem and its critics, what carries the load instead, and who counts the states. The ground Chapter 3 touches is wider than four pages, hence “eventually”. We would rather publish four pages we have checked than twelve we have not.
The part of this layer we care about most is the provenance marks. Each cheat sheet says where its knowledge comes from. ✓ means we have read the text. ◐ means the citation is confirmed but the content comes from an abstract or a secondary source. † means we have not read the source at all, and the mark stays visible. A page that flattered our reading would be cheaper to write, and it would not help you audit our work.
What we plan to achieve
For every theme we open we want four things: a claim, the context to read it without us, the neighbors to check it against, and a test that could go against it. The test is the receipt: the smallest piece of work that would settle the question, what it would cost, and what a negative result would show. Receipts are meant to collect in a public evidence register, where a claim that loses stays recorded as lost.
The context layer is our share. The rest needs other people.
Research opportunities
The list is longer than a blog post should be, so it has its own page: the questions about what a result would mean, the tests that need data we do not have, and the places where we suspect we are ignorant. If you are a researcher in one of these areas, a student looking for a topic, or someone who has tried this at work and kept records, one of the items is probably yours.
One request no literature check can meet: show us where we are ignorant. We mean the places where we do not know that we do not know, more than a wrong page number. A first public draft is the cheapest place to be embarrassed.
What we are asking for on accessibility
So far a theme has an introduction, a neighbor list and four deep dives. We do not know whether that is the right shape, and we would rather not guess what “accessible” means for you.
Send us a concrete proposal for making this material easier to reach. Some examples of what we mean: a graded reading order, one for someone who wants the whole argument and one for someone with ten minutes and one claim; a glossary of the terms we use as if they were obvious; one-page primers per theme; recorded walkthroughs of a chapter by its author; translations; a reading group that meets on the live text; an FAQ of the standard objections, with our answers. Some of these we can do, some we would need help with, and we will say which is which.
Say what your proposal would take and whom you would need. We will report back on what we take up, and on what we decided against and why.
How to take part
The Zotero library is open and holds most of what we cite. If you cannot get something, ask us. Each chapter page links to its discussion on Slack. The WhatsApp group is public, and the Buzz community is restricted to people who want to work on this. Or write to us directly.
Since it opened, this blog has kept one rule: older posts are not rewritten. When our understanding changes, a new post says so and links back to what came before. The manuscript runs the other way: a correction goes into the text with its date, so the trail stays visible.
Krishan Mathis and Margeret Heath