☞ New Types of Long-Term Scientific Institutions

I recently had a discussion with David Lang for Science Better about institutional experimentation in the sciences, the Overedge Catalog, and more (you can watch it here). This has been one of many recent discussions I’ve been having about the need for innovation in how we build research organizations.

And one question I’ve been thinking about is which aspects of institutional experimentation have we not seen yet in the wild. In other words, in the high-dimensional space of research organizations, which spaces are under-explored? One area seems to be institutional longevity. While more traditional institutions—particularly universities or philanthropic foundations—are not only built for the long-term but can last centuries, this doesn’t seem to be as true in the world of the organizations found in the Overedge Catalog. Of course, many of these organizations are very new, but I don’t think they are necessarily being built for long-term continuity, or even intergenerational persistence.

I’m affiliated with the Long Now Foundation, and as part of that, I’m involved in their Organizational Continuity Project, which is devoted to trying to discern the mechanisms for organizations lasting over long stretches of time. And related to this, I’m interested in the ways that we can build scientific institutions that will last. Because when these organizations last for the long-term, they can be better able to having enduring impact.

Of course, being long-lasting for long-term’s sake is not enough. An institution requires an enduring mission and vision, one that can sustain an organization multi-generationally. But if an organization can help articulate a clear vision and make sure that at each moment in time the people in that institution are working towards these goals, longevity can be a powerful multiplier.

And what are the mechanisms that might be necessary for such new types of research institutions? Obviously, a clear ever-green vision is vital. I also think a plan towards building and growing an endowment is an important feature. But there are many other open questions, from whether tenure-like job security is something that can aid longevity or hurt it, to whether being for-profit or nonprofit matters.

So, Dear Reader, I’d like to ask you: What are the mechanisms for long-term scientific institutions that the organizations of the Overedge Catalog should be experimenting with?

Bonus: Michael Nielsen has compiled a list of long-term projects, which includes some long-running scientific research projects.


I am pleased to announce that I’m going to be hosting my first Interintellect salon next month, focusing on how to think about technological complexity: Enlightenment to Entanglement: How to Think About Technological Complexity:

From machine-learning systems and desktop computers, to our automobiles and the infrastructure of the Internet, we are in an era of complex technologies, ones that are increasingly incomprehensible. And this is true not just for everyday users who might be confounded by the nature of smartphones and their computer’s software, but even for the experts who work with massive complex systems on a regular basis, or even built them. Regardless of our relationship with technology—users or builders—we require new ways of approaching these systems.

In this wide-ranging discussion, we’ll talk about everything from legacy code and AI to how our children should engage with technology to whether we have truly left the Enlightenment for the Entanglement.

I’d love to have readers join this conversation.

I recently came across tixy.land and was delighted. As readers know, I am intrigued by small bits of code that can unspool computational worlds. Tixy.land is a perfect example of what this kind of desire can result in. As per its tagline of “creative code golfing,” it is a fantastic combination of code golf and creative code: you write small snippets of code that can generate amazing visuals. Go play!

And a few other links worth checking out:

Until next month.

☞ NFTs and NP Problems

The Nature of Beauty meets Proof of Work?

In computer science, there is something known as an NP-complete problem: roughly, for this class of problem, finding the solution to such a problem is considered computationally difficult, but if I were to give you a potential answer, you could nearly instantly respond as to whether it was correct or not. For example, given a whole set of logical formulas, it might be hard to find a set of True/False values that satisfy all of them. But if I gave you a potential solution, you could easily check and see if my answer is the right one.

Inspired by Tony Kulesa, founder of biotech accelerator Petri, I’ve been thinking about NP-completeness recently in the context of NFTs. Tony had been intrigued by Robin Sloan’s amulets: Robin Sloan developed a method for identifying rare snippets of text, specifically ones that had certain properties when run through hash functions. The implication here is that the difficulty of finding rare bits of poetry—ones that have a hash value of four or more 8s in them—could be potentially mapped onto their inherent value.

When Tony and I were chatting about this, I was reminded of the idea of NP-completeness in terms of how we think about the problem of aesthetics. While it’s straightforward to know whether or not you like a specific poem or novel or painting—or digital artwork—it is much more difficult to create that specific work of art itself. Essentially, you can verify beauty, but it’s hard to generate it.

Now, as far as I know, there is no general-purpose algorithm for generating art. But if there were, it would be nice if the computational difficulty would map onto NP-completeness. In other words, as the “size” of the desired artwork grows, it would become much more computationally expensive, because art would remain rare and hard to find (though easy to verify as artistic). As a result of this, this would mean that the actual cost of an artwork could map onto its value. Would this finally provide a means of truly connecting artwork and NFTs?

Of course, a lot of this is no doubt ridiculous (and might be based on some fundamental crypto misunderstandings!), but I think there is value in thinking about how the blockchain, NP-completeness, and aesthetics contain some ideas that when mapped onto each other yield something thought-provoking.

Imagine a world where the energy expenditure of Bitcoin was instead being used to produce computationally-generated beauty…

Thanks to Tony Kulesa for discussions about this.

In Overedge Catalog news, I am interested in collecting unconventional revenue models for research organizations, such as research report subscriptions (such as Fast Forward Labs) or corporate membership/sponsorship (such as the Santa Fe Institute). If you know of interesting models, please pass them along.

I wrote a piece for The Atlantic about the recent Fastly internet outage: “Nobody Understands the Internet Until It Breaks

And some additional links worth checking out:

Also, alas, the signal from Proxima Centauri seems to just be interference.

Until next time.


☞ The Philosopher's Tree

Alchemy, Computation, and the Complexity of Modeling the World

The Philosopher’s Stone was an elusive goal of alchemists, ostensibly capable of producing the elixir of life and transmuting lead into gold. Or maybe it was itself the elixir of life? When it comes to the ideas in alchemy, I find things to be a bit hazy (it also doesn’t help that the Philosopher’s Stone doesn’t actually exist). But unlike the Stone itself, a presumed precursor for the Stone was actually able to be made: the Philosopher’s Tree.

The Philosopher’s Tree, which I learned about in Carl Zimmer’s new book Life’s Edge, was a tree-like crystal that could be “grown” under certain conditions. As per Wikipedia:

It is a dendritic amalgam of crystallized silver, obtained from mercury in a solution of silver nitrate; so-called by the alchemists, among whom “Diana” stood for silver. The arborescence of this amalgam, which even included fruit-like forms on its branches, led pre-modern chemical philosophers to theorize the existence of life in the kingdom of minerals.

Points for the use of the word “arborescence” aside, the philosopher’s tree reminds me of constructs found in the world of complexity science, especially what results from the process of diffusion-limited aggregation:

Diffusion-limited aggregation, or DLA, is the process of successive randomly-walking points aggregating together upon an original starting seed. And they are not that hard to program (I vaguely recall making one years ago using the programming language Processing).

The science writer Philip Ball has written a gorgeously illustrated book called Patterns in Nature, devoted to showing the categories of these patterns and exploring the reasons behind these ubiquitous kinds of shapes, whether cracks in a surface, fractal-like shapes, or much more.

And those in the computational realm have run with this approach, simplifying models down to their very bones and constructing computational toys powerful in their rich generative creativity, from cellular automata to falling-sand games and L-systems, each of which uses a small set of rules to generate complex and beautiful natural phenomena.

As someone who has spent lots of time in the field of complex systems, I am still entranced by this possibility (I’ve previously written about small chunks of code unspooling computational worlds). Perhaps there is even a hint of the alchemical wish here: that if we can find the set of rules and primitives, we can generate massive complexity.

And yet. The world is incredibly messy, and not everything—or even most of the world—can be fully understood by models like these. These are simplifications, powerful in their evocativeness and guides to our intuition, yet far from able to completely explain our world. L-systems can give us insight into plants, but they are not one and the same. So too with cellular automata and the patterns on shells. (This kind of complexity is discussed further in my book Overcomplicated.)

But that’s okay. As long as we realize what models are for—and how amazing and unifying they can sometimes be!—we can gain understanding and make progress. The Philosopher’s Tree never got us to the Philosopher’s Stone, but it might still teach us something about the complex field of chemistry.


I’ve been continuing to add organizations to the Overedge Catalog. If anyone has any further suggestions, please reach out. Also, the Catalog is on the verge of becoming too unwieldy for all entries to be listed on a single page, so if you have any suggestions for other modes of organization for the Catalog, please let me know.

Is the richness of the world embedded within GPT-3? Perhaps…

A few links worth checking out:

Until next month.

☞ The Overedge Catalog

Finding the misfit research + tech organizations

I recently created something called the Overedge Catalog, which I’m really excited about. Here’s the description from the site:

Research organizations and institutions often are shoehorned into a set of well-established categories: universities, public companies, tech startups, and certain types of non-profits, such as think tanks. But there is the need for innovation here, particularly when it comes to encouraging the development of new ideas and the ability to operate on long timescales. We need new types of research organizations.

In cartography, most maps are bound by the straight lines at their borders. But occasionally, there are parts of the map that don’t quite fit. They bleed over the edge and yet still cry out for being included in a map. These are the overedges. The Overedge Catalog is devoted to collecting the intriguing new types of organizations and institutions that lie at the intersection of the worlds of research and academia, non-profits, and tech startups. This is a small but growing number of organizations, but hopefully by collecting and highlighting all of these here, it can spur further institutional innovation.

I currently have about forty organizations included, along with a simple taxonomy, from Dynamicland and the Santa Fe Institute to Stripe Press and the Recurse Center.

Since I first started publicizing the Overedge Catalog a couple weeks ago, many folks have reached out with further suggestions (thanks so much to all of you!) and I have begun making more additions, with hopefully even more to come. Please check out the current version, share it with your friends, and most importantly, please provide any further suggestions or other feedback.


Related to my perennial love of exploring old computing magazines, Matt Webb pointed me to this incredible treasure trove: the junk mail of Ted Nelson. Amazing.

Also, I recently discovered a Palantir from the 1980’s in an issue of MacUser:

Never stop digging in the archives.

And here are a few quick links:

  • Wikipedia Speedruns: Finding the deep interconnectivity of Wikipedia as fast as you can.

  • GPT-3 tries pickup lines: These are bonkers and hilarious.

  • Radioactive data: tracing through training: “We want to detect whether a particular image dataset has been used to train a model. We propose a new technique, radioactive data, that makes imperceptible changes to this dataset such that any model trained on it will bear an identifiable mark.”

  • And Alan Jacobs explores what he finds most important: “The point, at this stage in my career, after fifteen published books, is not the publication, it’s the thinking. So let the thinking, in public, commence.”

Until next month.

☞ Timestamping Science and Magical Text

How do you demonstrate that you know something—or discovered something first—without giving away your secret? This was an issue that loomed large in the early years of modern science, as scholars wished to maintain priority without telling everyone what they had figured out, especially if it involved work that was still in progress.

So, when I was reading James Gleick’s biography of Isaac Newton recently (it’s fantastic, so please go check it out), I was intrigued by the following quotation:

5accdæ10effh11i4l3m9n6oqqr8s11t9y3x: 11ab3cdd10eæg10ill4m7n6o3p3q6r5s11t8vx, 3acæ4egh5i4l4m5n8oq4r3s6t4v, aaddæcecceiijmmnnooprrrsssssttuu

This odd string of letters and numbers was written by Newton in a letter, as a way of demonstrating via a sort of encryption that Newton had figured out various aspects of calculus.

How did this work? As explained here, “he used a simple procedure: he wrote a sentence (in Latin) and then just counted letters in it. And the anagram consisted of the list of letters and how many times each letter occurs in the message.”

And this is not the only example. In another of Newton’s letters, there’s also this string, “6accdae13eff7i3l9n4o4qrr4s8t12ux,” used to establish priority for another facet of calculus.

These long strings of letters and numbers remind me of the output of hash functions, like MD5 or SHA-2, which are often used as checksums, to ensure such things as a file having been transferred correctly. The problem with this anagram method though, is that it is much easier to reverse. While modern hash functions that are intended for cryptographic use are designed to be one-way—that is, you can’t figure out the original data from the function’s output—this needn’t be true for these anagram-based trusted timestamp approaches.

In fact, an attempt to reverse one of these actually led to an inadvertent discovery. Galileo had used an anagram to disguise his discovery of the bumpy shape of Saturn—which were its rings, though he wasn’t quite sure what it was and described it has having three parts—through the use of this string: SMAISMRMILMEPOETALEUMIBUNENUGTTAUIRAS

Fellow scientist Kepler, however, attempted to actually determine what the converted phrase said by reversing the anagram. And in a strange twist, he came up with an entirely different Latin phrase, but one that was actually correct: that Mars has two moons! These moons weren’t discovered though until more than two hundred years later (the entire story can be read about here).

I would like to think that we have come far since these anagram-mad days, though I still think that scientific publication could use more than a few updates. That being said, seeing these scientific hash functions reminds me a bit of the writer Robin Sloan’s recent experiments with NFTs to create text-based amulets: short bursts of text that have a particularly rare cryptographic hash, giving them a special sort of value.

Are these mad alphanumeric strings from hundreds of years ago perhaps a bridge then between alchemy and magic and our modern understanding of our world?

Abracadabra, please meet 6accdae13eff7i3l9n4o4qrr4s8t12ux.

And as a bonus, I made the connection between magic and mathematics yet more explicit with this short snippet of code that hashes text into a string of alchemical symbols from Unicode:

Welcome to my laboratory

is transmogrified into


Please feel free to play with the code for Base🜀.


I recently had the pleasure of participating in a wide-ranging conversation that touched on the complexity of technology, the nature of knowledge, and much more. Watch it here.

A few links worth checking out:

Until next month.

Loading more posts…