Constitution of Knowledge

June 22, 2021 • #

Jonathan Rauch’s latest book The Constitution of Knowledge just dropped, which sounds sort of like a sequel, or at least a redux of his classic Kindly Inquisitors.

Brookings held a panel on his book’s release with historian Anne Applebaum and novelist Neal Stephenson (yes, that Neal Stephenson). In Constitution he follows up his ideas on liberal science and free speech with further work on institutional decay, social coercion, and disinformation.

I wrote about Kindly Inquisitors and Rauch’s liberal science concept in Res Extensa #9: Liberal Science — Fallibilism and Error Correction. His work covers critical first principles that we’re gradually navigating away from.


June 21, 2021 • #

Last year I switched to Airr as my main podcast app when they launched the beta, and have used it exclusively just about every day since.

Airr’s killer feature is the “AirrQuote”, which lets you clip snippets of podcast audio to share. There’s no other podcast app like it with as many integrations, like highlighting and syncing to your Readwise knowledge management workflow. It also has transcripts for tons of shows, which is a feature I didn’t know I wanted til I tried using Spotify or Overcast again and couldn’t scan through the shows in text form.

An AirrSpace for [Norman Chella]('s excellent [RoamFM]( An AirrSpace for Norman Chella’s excellent RoamFM

Hot off the presses last week, the Airr team shipped a slick new feature called “AirrSpaces”, which adds an audio chat room sort of functionality, an interesting innovation in a podcast player. An AirrSpace is like a combination of a podcast, Discord group chat, and Clubhouse room — hosts record clips and post into the room, which can be played in top to bottom sequence, like a normal podcast conversation. Others in the room can submit clips, too, like questions or comments for the host to review and post to the room or reply to. The Airr team hosted an introductory space you can take a look at as an example.

All of that is a cool mixture of some existing ideas, but they’ve not been done before in a podcast app. Podcast creation has historically been surprisingly lacking in client apps. What Clubhouse, Twitter Spaces, and AirrSpaces are bringing to the party is participatory audio: lowering the barrier for creating and enabling direct audience involvement. One of my favorite features is the asynchronous nature of an AirrSpace. One can be active for days with new conversation to return to, just like entering a group chat and catching up.

The “aftershow” genre has been a pattern catching on in Clubhouse; perhaps AirrSpaces will create a form of audio comment threads per episode that can spin up a new AirrSpace for post-show discussion on each one.

Weekend Reading: Bubbles, the Magic of Hobbies, and Legitimacy

June 18, 2021 • #

🗯 Well-Behaved Bubbles Often Make History

Byrne Hobart wrote this piece in the inaugural edition of a16z’s new publication, Future. On bubbles and their downstream effects:

Bubbles can be directly beneficial, or at least lead to positive spillover effects: The telecom bubble in the ’90s created cheap fiber, and when the world was ready for YouTube, that fiber made it more viable. Even the housing bubble had some upside: It created more housing inventory, and since the new houses were quite standardized, that made it great training data for “iBuying” algorithms — the rare case where the bubble is low-tech but the consequences are higher-tech. But, even so, there’s always the question of price: how can you tell when it’s worth the hype?

💡 A Project of One’s Own

There’s something special that happens when you allow your kids to treat hobbies like serious endeavors instead of playtime or games. Paul Graham’s latest:

Instead of telling kids that their treehouses could be on the path to the work they do as adults, we tell them the path goes through school. And unfortunately schoolwork tends be very different from working on projects of one’s own. It’s usually neither a project, nor one’s own. So as school gets more serious, working on projects of one’s own is something that survives, if at all, as a thin thread off to the side.

It’s a bit sad to think of all the high school kids turning their backs on building treehouses and sitting in class dutifully learning about Darwin or Newton to pass some exam, when the work that made Darwin and Newton famous was actually closer in spirit to building treehouses than studying for exams.

My interests in history and tech trace straight back to my time in high school building computers to play Civilization II. Personal projects have long term benefit if nurtured.

The Most Important Scarce Resource Is Legitimacy

On the heels of finishing Schelling’s collection of essays on game theory, I read this piece from Vitalik Buterin on legitimacy, a force that underpins any successful coordination game, of which the world of cryptocurrencies and DAOs are prime examples.

In almost any environment with coordination games that exists for long enough, there inevitably emerge some mechanisms that can choose which decision to take. These mechanisms are powered by an established culture that everyone pays attention to these mechanisms and (usually) does what they say. Each person reasons that because everyone else follows these mechanisms, if they do something different they will only create conflict and suffer, or at least be left in a lonely forked ecosystem all by themselves. If a mechanism successfully has the ability to make these choices, then that mechanism has legitimacy.


June 17, 2021 • #

“Bring yourself back online…“

Bernard Lowe


When I pumped the brakes on my daily writing routine last year, I had designs on some other interesting projects to spend time on that the daily demand wasn’t giving me space for.

Throughout 2019 and 2020, I’d built a decent muscle for repetition and managing good habits through the accountability of publishing monthly reports on each goal. The first of each month I’d put together my stats on progression. I never shared them widely, but the act of putting it out there on the open web on a regular schedule created a forcing boundary to go through the motions of self-reflection. I was writing a post each day (with a fairly low bar for what constituted a post), running regularly to hit an annual miles target, meditating, and tracking the books I read.

In the fall I started writing Res Extensa, a newsletter project on some deeper themes, which is something I’d wanted to do for a long time on the blog, and occasionally did, but the need for the daily heartbeat of publishing didn’t give me the breathing room to spend much time on longer pieces. Time is precious for most of us, and for me it was all I could do to keep up with the goal commitments I made for myself, without trying to make additional promises about a weekly, biweekly, or hell, even monthly newsletter-writing schedule on top.

For 2021 I decided not to repeat my annual ritual of goal-setting in the same way that I’d done for ‘19 and ‘20. I started to lose steam and wanted to take a breather after I hit the two-year mark. In my November update from last year, I wrote how it was feeling like going through the motions rather than driven by excitement and motivation. To some degree that’s the whole point of accountability forcing functions like this for habit-forming: do the reps even when it’s not fun. But beyond the rep-fatigue of many goals, I’d wanted to try working on some new ideas. While I’m down with being aggressive in pursuit of goals, there’s a thin line between aggressive and overcommitted. Overcommitment results in poor performance on all fronts. Focus, by definition, requires fewer targets.

With reading I’d decided not to set a goal as I had the past 2 years, with a book count target for the year. For obvious reasons it’s sort of imaginary to quantify meaningful reading and learning through a raw count of books. If you measure to hitting 40 books per year, you might shy away from deep, challenging reads in favor of quicker ones just to hit a number. This is the nasty downside of bad measurement — you start performing to the measure rather than in service of an underlying goal. In this case, reading interesting things is the real mission; tracking a count is just a way to keep enough pressure on yourself to spend time on it1.

Running was a goal that I was doing mostly fine with from a time management perspective. Unlike writing or learning, the time input required scales linearly, so it’s easier to fit in. One’s fatigue level isn’t consistent — some days you’re exhausted and really don’t want to do the miles — but at least 30 minutes of time results in 30 minutes of running. 30 minutes of time writing doesn’t guarantee 30 minutes of actual, readable words! My issues with consistent exercise in 2021 so far have been more due to schedule mayhem than anything else. Buying the house and moving earlier this year, plus readjusting and getting settled resulted in not much time left for putting in the miles. I’ve started getting back to it the past few weeks, but exercise is an area I need hard targets for to push myself consistently.

So back to the reboot.

I’m getting back on the horse of writing regularly here, and planning to set some reasonable targets for other goals for the remainder of the year. My thought that eliminating the hard personal goal targets would make space for other things was logical, but paradoxically made me get less done, reducing motivation overall to work on any personal projects. I thought having more time available would make a biweekly newsletter pretty easy, but it’s done the reverse. There’s literature out there on this topic, and many of us have experienced this firsthand: having a compressed schedule of availability focuses our attention on the things that matter most. When we have too much time, things can happen “whenever”, which turns out quite often to be “never”.

In the middle of 2019 when I was hitting all my goals regularly, I don’t remember feeling overwhelmed at all. Time management was better. I wasn’t thrashing my time away with other meaningless activities telling myself “I’ll get to that article later”.

The shift away from hard goals was a worthy experiment to see how much my habit-forming tactics of the previous 2 years worked. It turns out a habit goes away when you stop doing it. I learned that the hard number staring me in the face expecting to be hit is an excellent motivator for me personally, whether I like it or not. It’s okay though, the point of it all is the delicious sausage at the end, not how the sausage gets made.

  1. Inspired by Julian Lehr’s quantified self system, I started to track time spent reading as an alternative to book count, which is actually a more objective-aligned number to work from. Something worth talking about in a future post. I’m nowhere near as advanced as Julian on this, just using a system like this for a few things. 

M1 Mac Mini

June 15, 2021 • #

I just got a new Mac Mini with the M1 Apple silicon.

The experience so far is stunning performance compared to my previous 16” MacBook Pro. I was using an i9 with 16GB RAM, and this Mini blows it out of the water on responsiveness (and every other category).

M1 Mac Mini

A little reading on user experiences with the M1 had me interested in upgrading to any machine with the latest SoC. One of my main drivers was the noise and heat generated by the MBP, which is just in constant turbo mode with whatever my usage behavior is. It never stops running full tilt basically, so I needed to get away from that. My office is in the corner of the house and doesn’t get great HVAC coverage with the door closed, so between that and the west-facing windows, the heat-radiating laptop can’t have helped.

With the M1 Mini and a nice USB-C dock with a built-in fan that it sits on top of, I haven’t heard a sound from the machine at all. 11/10 so far. It’s wild that such an affordable, portable desktop machine has owned everything pre-M1 in performance.

DeFi Explainers

May 4, 2021 • #

I’ve gone over off the deep end the last couple weeks trying to wrap my head around DeFi. To date I’ve only dabbled in crypto, being lucky enough to ride some small waves, though nothing life-changing.

DeFi (decentralized finance) is fascinating for its disruption potential (and Ethereum platform on which it’s all built). A basic understanding of the conceptual possibilities shows this stuff is here to stay, even if not in the same form or as loud as meme-ish as it’s been over the past year.

Through Twitter I discovered a channel called Finematics that has a ton of great explainer videos walking through topics like smart contracts, the history of DeFi, NFTs, yield farming, and dozens of others on the esoteric crypto world.

Check out the full channel for the archive.

On Legibility — In Society, Tech, Organizations, and Cities

April 6, 2021 • #

This is a repost from my newsletter, Res Extensa, which you can subscribe to over on Substack. This issue was originally published in November, 2020.

In our last issue, we’d weathered TS Zeta in the hills of Georgia, and the dissonance of being a lifelong Floridian sitting through gale-force winds in a mountain cabin. Last week a different category of storm hit us nationwide in the form of election week (which it seems we’ve mostly recovered from). Now as I write this one, Eta is barreling toward us after several days of expert projections that it’d miss by a wide margin. We’re dealing with last-minute school closures and hopefully dodging major power outages. 2020 continues to deliver the goods.

There’s a lot in store for this week, so let’s get into it:

Seeing Like a State

I’ve been deep in James C. Scott’s Seeing Like a State lately, so I wanted to riff this week on Scott’s notion of “legibility,” the book’s central idea. His thesis in SLAS is that central authorities impose top-down, mandated designs on societies in order to make them easier to understand through simplification and optimization techniques. Examples in the book range from scientific forestry and naming traditions to urban planning and collectivist agriculture.

Boca Raton's legible landscape

Scott calls this ideology “authoritarian high modernism,” wherein governments, driven by a zealous belief in knowledge and scientific expertise, determine they can restructure the social order for particular gain: higher crop yields, more compliant citizenry, more efficient cities, or crime-free neighborhoods (a dangerous proposition when a “crime” is redefined at will by authorities). He’s ruthlessly critical of these ideas, as evidenced by the book’s subtitle, and presents dozens of cases of top-down-design-gone-wrong, the most extreme case being the Soviet Union’s collectivization program that led to widespread famine.

An interesting factor to think about is how and when to apply intentional design in service of legibility and control. It’s not an all good–all bad proposition, to be sure. Even though Scott levels a pretty harsh review of high modernist ideology, even he acknowledges its value in small, targeted doses for specific problems.

Legibility: A Big Little Idea

I linked to a piece a while back by Venkatesh Rao, the source where I first learned of Scott’s work.

The post is largely an introduction to the book’s themes, but adds a few interesting notes on the psychology behind legibility. Given all the history we have that demonstrates the failure rate of high modernist thinking, why do we keep doing it?

I suspect that what tempts us into this failure is that legibility quells the anxieties evoked by apparent chaos. There is more than mere stupidity at work.

In Mind Wide Open, Steven Johnson’s entertaining story of his experiences subjecting himself to all sorts of medical scanning technologies, he describes his experience with getting an fMRI scan. Johnson tells the researcher that perhaps they should start by examining his brain’s baseline reaction to meaningless stimuli. He naively suggests a white-noise pattern as the right starter image. The researcher patiently informs him that subjects’ brains tend to go crazy when a white noise (high Shannon entropy) pattern is presented. The brain goes nuts trying to find order in the chaos. Instead, the researcher says, they usually start with something like a black-and-white checkerboard pattern.

If my conjecture is correct, then the High Modernist failure-through-legibility-seeking formula is a large scale effect of the rationalization of the fear of (apparent) chaos.

Scott also points out in the book how much of the high modernist mission is driven by “from above” aesthetics, not on-the-ground results. He analyzes the work of Le Corbusier and his visionary model for futuristic urban planning, most evident in his designs for Chandigurh in India and the manufactured Brazilian capital of Brasília (designed by his student, Lúcio Costa).


While Brasília projects a degree of majesty from above, life on the street is a hollowed-out, sterile existence. Its design ignored the realities of how humans interact. Life is more complex than the few variables a planner can optimize for. It’s telling that both Chandigurh and Brasília have lively slums on their outskirts, unplanned neighborhoods that filled demands unmet by the architected city centers.

He juxtaposes the work of Le Corbusier with that of Jane Jacobs, grassroots city activist and author of The Death and Life of Great American Cities. Jacobs was a lifelong advocate of “street life,” placing high emphasis on the organic, local, and human-scale factors that truly make spaces livable and enjoyable. She famously countered the ultimate high modernist visions of New York City planner Robert Moses.

This contrast between failure in legible, designed systems and resilience in emergent, organic ones triggers all of my free-market priors. The truth is there is no “best” mental model here. Certain types of problems lend themselves well to top-down control (require it, in fact), and others produce the best results when markets and individuals are permitted to drive their own solutions. Standard timekeeping, transportation networks, space exploration, flood control — these are all challenges that are hard to address for a variety of reasons without centralized coordination.

Imposing legibility demands an appreciation of trade-offs. Yes, dictated addressing schemes and fixed property ownership documentation do enable state control in the form of taxation, conscription, or surveillance. But in exchange for the right degree of imposed structure, we get the benefit of property rights and land tenure.

Big Tech’s Legible Vision

Byrne Hobart touched on this idea in an issue of his newsletter. (The Diff is some of the best tech/business/investing writing out there, I highly recommend subscribing). He makes the point that wherever scale is required, abstraction and legibility are highly valuable. He calls up Scott’s usage of the term mētis, translated from the classical Greek to mean roughly “knowledge that can only come from practical experience.” The value of this local knowledge is Scott’s counter to legibility-seeking schemes: that practical knowledge beats the theoretical every time.

I like this point that Hobart makes, though, on seeking larger scale global maxima: that a pure reliance on the practical can leave you stuck in a local maximum:

But mētis is a hill-climbing algorithm. If it’s based on experience rather than theory, it’s limited by experience. Meanwhile, theory is not limited by direct experience. By the 1930s, many physicists were quite convinced that an atomic bomb was possible, though of course none of them had ever seen one. Because some things can’t be discovered by trial and error, but can be created by writing down some first principles and thinking very hard about their implications (followed by lots of trial and error), the pro-legibility side has an advantage in inventing new things.

He also brings up its implications in the modern tech ecosystem. The Facebooks, Googles, and Amazons are like panoptic overseers that can force legibility even on the most impenetrable, messy datastreams through machine learning algorithms and hyper-scale pattern recognition. The trade-off here may not be conscription (not yet anyway), but there is a tax. There’s an interesting twist, though: the tech firms have pushed these specialized models so far to the edge that they themselves can’t even explain how they work, thereby reconstituting illegibility:

Fortunately for anyone who shares Scott’s skepticism of the legibility project, the end state for tech ends up creating a weird ego of the mētis-driven illegible system we started with. The outer edges of ad targeting, product recommendations, search results, People You May Know, and For You Page are driven by machine learning algorithms that consume unfathomable amounts of data and output a uniquely well-targeted result. The source code and the data exist, in human-readable formats, but the actual process can be completely opaque.

Functional versus Unit Organizations

While legibility interests me in its applicability to society as a whole, I’m even more intrigued by how this phenomenon works on a smaller scale: within companies.

Org charts attempt to balance productivity and legibility, which often pull in different directions. Organizational design is driven by a hybrid need:

  1. To ship products and services to customers in exchange for revenue and equity value, and
  2. To be able to control, monitor, and optimize the corporate machine

Companies spend millions each year doing “reorgs,” often attributing execution failures to #2: the illegibility of the org’s activities. Therefore you rarely see a reorg that results in drastic cutback of management oversight.

Org Chart

Former Microsoft product exec and now-VC Steven Sinofsky wrote this epic piece a few years back comparing the pros and cons of function- and unit-based organizational structures. Each has merits that fit better or worse within an org depending on the product line(s), corporate culture, geographic spread, go-to-market, and headcount. The number one objective of an optimum org chart is to maximize value delivery to customers through cost reduction, top-line revenue gains, lower overhead, and richer innovation in new products. But legibility can’t be left out as an influence. The insertion of management layer is an attempt to institute tighter control and visibility, a degree of which is necessary to appropriately dial-in costs and overhead investment.

Look at this statistic Sinofsky cites about the Windows team’s composition when he joined:

One statistic: when I came to Windows and the 142 product units, the team overall was over 35% managers (!). But the time we were done “going functional” we had about 20% managers.

Even corporate teams aren’t immune to the pull of legibility. When its influence is stretched too far, you end up with Dilbert cartoons and TPS reports.

Emergent Order in Cities

It was serendipitous to encounter this same theme in Devon Zuegel’s podcast, Order Without Design. The show is a conversation with urbanist Alain Bertaud and his wife Marie-Agnes, an extension of his book on urban planning and how “markets shape cities.”

In episode 3 they discuss mostly sanitation and waste management in cities around the world, but my favorite bit was toward the end in a discussion on how different cities segment property into lots and dictate various uses through zoning regulation. Some cities slice property into large lots, which leads to fewer businesses and higher risk for large, expensive developments. But others, like Manhattan, segment into smaller chunks, resulting in a more diverse cityscape that mixes dining, retail, services, and many other commercial activities.

The shopping mall was an innovation that allowed developers with limited local knowledge to have tenants respond to customer demand in smaller, less risky increments. Malls in Asia notably differ from ours in the west in the breadth of goods and services they typically offer. Here’s Marie-Agnes from the episode:

…Remember the Singapore mall. In Singapore the malls have not only retail, but you find also dental offices, notaries, kindergarten schools, dispensaries, and all sorts of activities. Up to now and in general the concept of a mall in the US is for mainly shopping. You may have food court and some restaurants, but nothing like a real city where you have all sort of business.

Devon points out that this quality might make these malls more resilient to market stress than our retail-focused American versions.

Malls are by no means a modern innovation, of course. Ad-hoc congregations of commercial activity have been around for 5,000 years, evolving from purely organic bazaars of the Near East into the pseudo-planned, air conditioned behemoths we have today. Even in our technocratic culture, planners still realize that the flexibility to market demand is crucial to sustainability. Trying to pre-design and mandate a particular distribution of stores and services is a fool’s errand.

In Byrne’s newsletter, he says “once you look for legibility, you start to see it everywhere.” This has certainly been true for me as I’ve been reading Scott’s work. The deepest insights are the ones that cut widely across many different dimensions.

There’s no silver bullet in how to apply legibility-inducing schemes to any of these areas. But until you’re aware of the negative consequences, there’s no way to balance the scales between legible/designed and illegible/organic outcomes.

The Low-Code IKEA Effect

March 22, 2021 • #

I linked a few days ago to Packy McCormick’s piece Excel Never Dies, which went deep on Microsoft Excel, the springboard for a thousand internet businesses over the last 30 years. “Low-code” techniques in software have become ubiquitous at this point, and Excel was the proto-low-code environment — one of the first that stepped toward empowering regular people to create their own software. In the mid-80s, if you wanted to make your own software tools, you were in C, BASIC, or Pascal. Excel and its siblings (Lotus 1-2-3, VisiCalc) gave users a visual workspace, an abstraction layer lending power without the need to learn languages.

Today in the low-code ecosystem you have hundreds of products for all sorts of use cases leaning on similar building principles — Bubble and Webflow for websites, Integromat and Zapier for integrations, Notion and Coda for team collaboration, even Figma for designs. The strategy goes hand-in-hand with product-led growth: start with simple use cases, be inviting to new users, and gradually empower them to build their own products.

Low-code IKEA effect

Excel followed this model since the 80s: give people some building blocks, a canvas, and some guardrails and let them go build. Start out with simple formulas, create multiple sheets, cross-link data, and eventually learn enough to build your own complete custom programs.

What is it about low-code that makes it such an effective strategy? Sure there’s the flexibility it affords to unforeseen use cases, and the adaptability to apply a tool to a thousand different jobs. But I think there’s psychology at play here that makes it particularly compelling for many types of software.

There’s a cognitive phenomenon called the “IKEA effect”, which says:

Consumers are likely to place a disproportionately high value on products they partially created

IKEA is famous for its modular furniture, which customers take it home and partially assemble themselves. In a 2011 paper, Michael Norton, Daniel Mochon, and Dan Ariely wrote identified this effect, studying how consumers valued products that they personally took part in creating, from IKEA furniture, to origami figures, to Lego sets. Other studies of effort justification go way back to the 1950s, so it’s a principle that’s been understood, even if only implicitly, by product creators for decades.

Low-code tools express this effect, too, showing that customers are very willing to participate in the creation process if they get something in return. In the case of IKEA it’s more portable, affordable furniture products. In low-code software it’s a solution tailored to their personal need. Paradoxically, putting additional effort into a product through self-service, assembly, or customization generates a greater perception of value than consumer being handed an assembled, completed product.

SaaS companies should embrace this idea. Letting the customer take over for the “last mile” can be wins all around for everyone. Mutual benefits accrue to both creator and consumer:

  • Customers have a sense of ownership when they play a role in building their own solution.
  • The result can be personalized. In business environments, companies want an added touch of differentiation, to set themselves apart from competitors. They don’t want commodities that any other player can simply buy and install.
  • Production costs are reduced. The creator builds the toolbox (or parts list, instructions, and tool kit) and lets the consumer take it from there. Don’t have to spend time understanding the nuances of hundreds of different use cases. Provide building blocks and let recombination generate thousands of unique solutions.
  • Increased retention! Studies showed that consumers consistently rated products they helped assemble higher in value than already-assembled pieces. This valuation bias manifests in retention dynamics for your product: if customers are committed enough and build their own solution, they’ll more likely imbue it with greater value.

The challenge for creators is to balance that “just-right” level of customer participation. Too much abstraction in your product, requiring too much building of primitives, and the customer is unlikely to have the patience to work through it. Likewise, when you buy an IKEA table, you don’t want to be sanding, painting, or drilling, but snapping, locking, and bolting are fine. Success is a key criteria to get the positive upside. From the Wikipedia page:

To be sure, “labor leads to love only when labor results in successful completion of tasks; when participants built and then destroyed their creations, or failed to complete them, the IKEA effect dissipated.” The researchers also concluded “that labor increases valuation for both ‘do-it-yourselfers’ and novices.”

Participation in the process creates a feedback loop: the tool adapts to the unique circumstances of the consumer, functions as a built-in reward, and the consumer learns more about their workflow in the process.

Low-code as a software strategy allows for a personalization on-ramp. Its IKEA effect gives customers the power to participate in building their own solution, tailoring it to their specific tastes along the way.

Subscribe here to receive my newsletter, Res Extensa, a digest of my latest posts, links, and updates. Currently bi-weekly.