The web3 world computer is at a 1970 level of development
20 May 2022 | 1:12 pm

I’ve been hanging out in the web3 space recently. The art and aesthetic generally is awesome and effervescent. But there are big claims made for the technology and, given the high level of scams, I’ve spend a bunch of time considering that.

With any new tech, it’s interesting to ask: does this matter? Is it a weird blip or does this become part of the technology landscape for decades to come? If so, how? A change in consumer expectations or wildly disruptive at a widespread and technical level? Does this tech matter directly, or is it basically a discovery mechanism for new use cases, acting as inspiration for new tech that answers that same revealed use case but in a different way?

Most interesting: where are we in computing history?

For e.g.: My take is that VR is waiting for its Macintosh moment which puts it in the mid 80s and my gut says that’s about right. Assuming decent smart glasses drop this year or next, then it’ll take a decade of deployment phase till we’re in the “mid 90s” and VR is transforming everyday life. At which point we’ll be ready for the big twist, which for the PC was the arrival of the web, and it took another decade for PCs to become culturally dominant. Perhaps, v roughly, VR is on the same curve.

Ok, web3.

So I’ve been poking around in Ethereum which is one of the two big blockchains. Bitcoin is the other and that’s mostly financial. Ethereum is… something else.

Reading the smart contract source code that underpins NFTs (here’s the spec and here’s some code) is super informative! You can see what it means for a digital item to “exist.” You can see what it means for one of these items to be “owned.”

I get the same feeling as when I read the source code for the original Unix operating system, which is basically the ur-OS that either directly or indirectly (because it established the concepts) underpins this epoch’s computing environment. You can see what a process is! What is a file is. What a user is. The feeling is a combination of: oh now I know the fundamental particles; and, is that it??

NFTs don’t “exist” on the Ethereum blockchain. The smart contract that tallies and tracks ownership for a class of NFTs does exist: it’s code that runs. As a single instance, this means that NFTs of the same class are forever linked. By analogy: imagine a print of some art is limited to 100 editions and they are all sold and now hang in people’s homes all over the world. Now imagine that they are spookily connected to one-another.

An NFT is a smart contract that achieves something like digital ownership, but actually it has a subtly different nature. Which is worth knowing. Opportunities for invention come from knowledge of the deep physics of a universe.

Long story short, I finally came to understand whey they call Ethereum the World Computer.

NFTs are one type of smart contract that can run on the world computer. Other smart contracts can do anything that code can do.

Smart contracts aren’t contracts, that’s a financial or legal framing. Smart contracts are object instances, in the object-oriented code sense, and the Ethereum blockchain is a shared object runtime.

Robin Sloan gets it. I didn’t get it when I read his notes last year, but now I do.

Ifeel like this simple premise is often lost in the haze: the Ethereum Virtual Machine, humming heart of Web3, is a computer that charges you many dollars to execute a very small program very slowly. It does so in an environment with special properties, and in some cases, those properties are worth the expense. In others … it’s like running your website on a TRS-80 with a coinslot.

(And also read Sloan’s essay from around the same time, The slab and the permacomputer, which is a meditation on the idea that "‘computers’ might melt into ‘compute’", something more environmental then physical.)

There is a picture in my head that the computing environment of the future is a vast shared substrate where digital ownership is IRL-equivalent. With all that ownership implies and requires: identity, economics, object permanence, a truth grounded deep in the physics.

Maybe achieving that requires tearing up everything way back to… well, when?

Let’s say that the Ethereum virtual machine, the world computer, is indeed the shared object runtime that we will need. Object-oriented code being the paradigm invented by Alan Kay such that blobs of code sit together, an object instance representing (say) an on-screen menu, or a user, or whatever. The paradigm provides abstraction such that code can reach dizzying complexity, while also retaining expressiveness to create new things.

Object runtimes such as…

  • NOW: the object-oriented heart of MacOS and iOS, originally the Objective-C runtime of NeXT (and the reason that Apple bought NeXT), which is the reason why iPhone has great consistency in the user experience and was also, at release, hugely easier to develop for than anything else on the market.
  • BACK THEN: the Smalltalk runtime, developed by Alan Kay’s team at Xerox PARC, and used to invent modern graphical user interface on the Xerox Alto in the early 1970s.

(Yes, I continue to be obsessed with the career history of Alan Kay.)

It’s interesting to me that you need both (a) the hardware, and also (b) an expressive and powerful object runtime such as Smalltalk if you are going to invent the user interface layer, which allows users to interact with the machine and also provides a platform for apps.

If I kinda squint… I can kinda imagine how you might bootstrap today’s Ethereum virtual machine all the way up to a Smalltalk-equivalent. And if you get to that point, you can ladder your way up to a GUI analogue, and from there to modern-day computing.

So how far away are we?

How far away is today’s web3 from something as sophisticated as today’s computer world?

Let’s do some sums.

And really hand wave our way to a Fermi estimation.

Executing a function on a smart contract (an object instance) on the Ethereum virtual machine, updating internal state etc, takes a few minutes.

On a Mac, the overhead to pass a message to an object is measured in nanoseconds on a modern machine.

So there’s is a 10 orders of magnitude difference: Ethereum needs to be 10 billion times faster.

That’s 33 Moore’s law doublings. 50 years away from being as complex and fully-expressed as today.

So we’re in the equivalent of 1970 – which feels about right.

Web3 is waiting for minicomputers. Even that’s a long way off from today. In the minicomputer boom around 1980, in our history, a single NAND gate cost 8 cents, wholesale: In 1981 money, a single iPhone would cost "$1.4 billion in parts, no margin."

We’re still waiting for our Unix moment, locking down the fundamental concepts, the system that takes the network and time-sharing for granted, giving us the native programming language and system calls to bootstrap up to the next layer of emergent complexity.

We’re pre GUI; direct manipulation and the desktop metaphor has yet to be figured out. There are no SDKs. Development is still close to the metal.

It gives me a rough handle on the scale of work to be done.

1970 doesn’t mean that web3, this new epoch of computing (if that’s what it is), is unusable. Far from it. People in the real 1970 were making video games! The personal computer had already been imagined and prototyped!

This analogy helps me have a view on questions like: are NFTs the final form of that concept, or do we have some way to go to digital ownership?

I think what NFTs want to be like is the MP3. The Fraunhofer Institute’s invention of the MP3 (and their licensing approach) unlocked a whole industry including consumer ownership of digital music, online music stores and streaming, and digital devices like the iPod (which paved the way to the iPhone).

But the MP3 file format was invented in 1989 so - if we’re on a similar trajectory - then NFTs have conceptually the right frame but are 20 years too early.

The question is: how do we accelerate 50 years to 20 years or to 10 years?

There were multiple generations of computers between 1970 and the networked smartphone. It wasn’t a steady evolution.

Maybe there would be scope in imagining the next generation of web3, already. Are there other ways to achieve a global, zero-trust, persistent, shared object runtime – and can it be built? Perhaps Microsoft or Google have warehouses of genius engineers doing just that, attempting the generational leapfrog.

Or maybe it would be worth bullying a Smalltalk-like expressive development environment into existence, sitting atop today’s Ethereum world computer, however slowly it run, just to see what could be created with that new clay.

IF REAL! The alternative view is that web3 and all of the above isn’t real. There will be ways to achieve digital ownership (if that’s even important!) without baking it into the physics of a future world computer.

It could be that part of the appeal of web3 is that it’s a new glass bead game.

(Whether it has 1,000 year appeal like THE Glass Bead Game, Hermann Hesse’s abstract and beautiful fictional game at the heart of his 1943 novel, our descendants will find out. So let’s lowercase it for now.)

Where else can you manipulate a novel set of symbols and bounce between code, social dynamics, arts and economics? There are endless permutations.

And perhaps - per Hesse - it’s best left to a caste of esoteric monks revered yet safely isolated from the rest of society…

I’m not saying there is nothing else there, or that the nerd-sniping joy of web3 is the only value (I happen to believe there is something real here), but alongside the gambling drive which comes from the financial component of this emerging tech stack, I feel like novel symbol manipulation is a big part of the early appeal for many.

Which is a risk. I kinda vaguely feel like physics burnt a couple decades on a similar pursuit: string theory. A consumingly absorbing idea for whole communities, but where did it go?

So perhaps it’s all a mirage. For the sake of argument, let’s assume it’s not.

My hunch and my heuristic metaphor:

If the web3 world computer has only just reached 1970 then, first, don’t expect too much. There’s real utility to be found but in very prescribed use cases. But also, second, there are wild and unrecognisable transformations to come. There is room for imagination and invention.

Lightbulbs were so startup
18 May 2022 | 6:40 pm

Edison invented the first practical incandescent electric light in 1879. (That is, the filament lasted more than a few hours.)

But it couldn’t be brought to consumers because there was no commercially-available electricity.


In 1882, the Edison Illuminating Company opened the first commercial power plant in the United States: Pearl Street Station (Wikipedia) in Lower Manhattan.

Starting small… "it started generating electricity on September 4, 1882, serving an initial load of 400 lamps at 82 customers. By 1884, Pearl Street Station was serving 508 customers with 10,164 lamps."

They had to invent a new kind of dynamo to generate the electricity. Ahead of the power station, Edison ran a number of pop-ups as prototypes.

HOWEVER: power distribution.

… perhaps the greatest challenge was building the elaborate network of wires and underground tubes (called “conduits”) needed to deliver energy to customers. New York City politicians were initially skeptical and rejected Edison’s proposal to dig up the streets of lower Manhattan to install the needed 100,000 feet of wiring. Eventually, however, Edison was able to convince the mayor of the city otherwise. The conduit installation proved to be one of the most expensive parts of the whole project.

Not only a distribution problem, but a backchannel conversation between people with power to get around the rules. Capitalists gonna capitalise.

AND THERE’S MORE: the business model.

Since the early 1800s there had been special instruments to detect the flow of a current and indicate how much of it was flowing, but there was not an instrument to record that flow over time. Not until the spring of 1882 was a successful design for an electric meter available. However, Edison did not send bills to his customers until the whole system was running reliably, which took some more time. The first electric bill was sent to the Ansonia brass and copper company on 18 January 1883 and was for $50.44.

(That quote from the same ETHW article linked above.)

In addition light bulbs cost $1 ea.

The setup cost (including real estate) was $300,000 – a lot to return, 50 bucks at a time.

The company ran at a loss until 1884. Pearl Street Station burnt down in 1890, was rebuilt, then in 1895 decommissioned when it was made obsolete by newer, larger power stations.

A single value-creating innovation requiring a vast hinterland of enabling technologies in order to connect the product to its market.

It’s so startup it hurts.

Questions I have:

  • How can we compare the cost of “inventing the lightbulb” vs digging the streets, building the power station, getting to profitability, etc? Say, in terms of hours of human effort? Was it 50:50, 80:20, 1:99?
  • Were other models other than metered electricity considered? Like, light-as-a-service with bundled bulbs? Or somehow connecting the cost of the service to the value for the customer – could customer businesses be charged a fraction of revenue, say? My guess: it comes down to what’s easy to measure. Businesses nowadays will grab a % (the “take”) from their customers simply because it’s possible to do so.
  • Were other applications for electricity actively considered, or was this a pure lighting play? The fractional horse-power motor scaled down factory automation to the home, and the electrification boom brought with it vacuum cleaning and washing machines… yet the enabling tech wasn’t invented till 1888. But was it visible on the horizon? Or was it only lighting that justified the huge investment in Pearl Street Station and the 100,000 feet of conduits?
  • Why go for this very broad market first? Why not build smaller generators in the basement of department stores, lighting up just single buildings or single streets? Isn’t the value in (say) adding an extra hour of brightly-lit shopping every evening more obvious than trying to replace gas and oil for home customers?

Did Edison have a team of 1880s MBA-equivalents, crunching the numbers to figure out what to do?

What was the mood around electric lighting back then? Did it feel like a hype train? Was the social media of the time full of wild speculation about the social changes that would be unleashed and the fortunes that would be made?

Electricity, generally, had that aura of excitement. A few years back I read every issue of Electrical Review from the 1880s and 1890s which covered the rollout of the telegraph and then lighting, simultaneous with figuring out the science of how electricity behaved (I wrote up my observations here) – but I think I need to go back and read the preceding decade too.

Signs of a magnetic pole flip in company ownership
12 May 2022 | 4:13 pm

What if the dominant model of company ownership inverts? What if we’re at the end of an era of companies being owned by external stockholders, and at the beginning of bottom-up ownership by the people who do the work – the employees? Feels unlikely I know, HOWEVER:

This morning’s news is that ustwo is now employee-owned.

You’ll likely know ustwo. Here’s their Wikipedia page. They’re behind the hit puzzle game Monument Valley; long-time digital design agency (founded in London in 2004) with a couple hundred staff; part of many joint ventures to provide design/software/marketing/etc for startups, e.g. DICE. I know Mills (@millsustwo), one of the two founders, from the general scene - huge congrats mate, brilliant move.

I have a soft spot for an ancient bit of ustwo work, being home screens designed for the Sony Ericsson XPERIA X2 (bloody hell that’s a mouthful) smartphone from 2009. Watch the Pixel City video on Vimeo: "Pixel city moves through a cityscape, its different elements linking to the functionality of your phone. Text messages appear playfully on billboards, calendar events arrive by train, a passing aeroplane shows your call history and much more."

ustwo have always been as inventive and pioneering with their business model as their design work. They’ve been blogging today about going employee-owned:

  • How: a majority of equity is now in a vehicle called an “Employee Owned Trust,” purchased from the founders with a loan (like a mortgage) which will be paid back over 6 years. Three employees have joined the board. There has been a multi-year transition leading up to this from being founder-led to management-led.
  • Why: employee-owned firms are "more productive, grow faster, and are less likely to go out of business" (where ownership is >30%; ustwo is now 62% employee-owned). There has always been profit share; now this can grow.

It’s great that they’re sharing the nuts and bolts of how this works. It’ll demystify the process for others who want to follow the same path.

And I’m sure there will be a bunch of future lessons in how to make this work – like, how can there be meaningful employee involvement in how to chose work or influence big bets or what happens when there are lean cycles in the agency cycle? I hope ustwo’s sharing will continue.

A shout-out at this point to my friends at Clearleft! A smaller but also well-established agency and extremely well-regarded for their design work and community presence, the Brighton-based design studio went employee-owned in 2020.

Two makes a trend right?

Exciting times for design. And for the agency model, which has been in a state of perpetual reinvention for as long as I remember.

RELATED: There’s something fascinating in thinking about succession planning as the founders handing control not to another individual to the machine. It makes me think of Sikhism which, after a line of 10 gurus, handed over leadership to an “eternal living guru,” the Sikh community itself. As previously discussed.

There’s always been the question about how founders exit from an agency. Two traditional routes:

  • Management buy-out funded by private equity. Going the PE route means “unlocking value” – there are strong growth expectations for a future flip (i.e. a return for the PE firm).
  • Acquisition by another agency, ideally a major network.

The agency world has its own nature and own norms – it’s like a more established, parallel world to the startup ecosystem.

One feature is the presence of behemoth networks like WPP plc (that’s their Wikipedia page). There are a handful of agency networks around the world. WPP is UK-based and owns a few hundred agencies, with collectively 100,000 employees and somewhere north of 10 billion annual revenue. They coordinate, share work, get scale (a small agency can be part of a global project), and save on back-office.

So while I love that ustwo and Clearleft are figuring out the path to employee-ownership – the eternal living guru of the organisation…

…thinking about the larger scale makes me ASK:

What is the equivalent of the agency network for employee-owned orgs?

Can we imagine some kind of multinational network organisation that coordinates, shares work, achieves scale, etc, without taking full control of the member agencies?

Going further:

The agency world has a fractal structure. Agencies are often 50% freelancers and they subcontract like crazy. Then they roll up into bigger firms – the Coasean logic of travel towards lower internal transaction costs means agencies combine and combine again until you get the network giants.

(The startup world parallel is the data-driven gravitational force which results in Big Tech, a.k.a. Srnicek’s platform capitalism.)

Can employee-ownership exist at all scales?

Hey and here’s an example in the UK! CoTech is a network of 45 creative technology companies, all individually organised as co-ops, providing digital services together. More like that pls.

(Thank you to the folks at the co-op Common Knowledge for letting me know about this. Common Knowledge itself creates digital tools to force-multiply social movements.)

Perhaps we’re at the beginning of an ownership inversion where organisations from big to small will follow the principle of bottom-up agency.

Dominant models change every so often! I remember reading that the dominant model in the US relatively recently (1900s?) was family-owned businesses. I’ll have to hunt down that reference.

The analogy here is to geomagnetic reversal, the process by which the Earth’s magnetic poles flip – the North Pole becomes the South Pole; the south becomes the north. It happens periodically: "There have been 183 reversals over the last 83 million years (on average once every ~450,000 years). The latest, the Brunhes-Matuyama reversal, occurred 780,000 years ago."

(i.e. we’re overdue, just in case you were wondering what else the 2020s may deliver.)

So I guess something happens such that the pole-flip kicks off, and sheer magnetism drags everything else with it to complete the process? There is no halfway house.

Once I started looking for signs of an ownership inversion, labour becomes capital and capital becomes labour, then I started seeing it everywhere:

  • Unionisation – it’s growing in tech and service jobs, and not directly employee-ownership but still an agitation for self-determination.
  • Independent employees – Uber drivers, Airbnb hosts, and the gig/sharing economy generally… are these employees stripped of rights, or individuals exerting self-determination, or (as I’ve previously argued) a third category of worker, something in-between? Maybe there’s the seed of employee-ownership hidden here. Let’s roll in the creator economy here: as Uber drivers love their flexibility, creators love the hustle.
  • Co-opsThe Drivers Cooperative in New York, the Uber-alike where the drivers own the software, and others in the platform co-op movement: organisations where management is the commoditised class, not the workers.
  • DAOs (Decentralized Autonomous Organisations), looking further out into the web3 world – a way for the bureaucracy of an organisation to be represented as executable code, and for individuals and other individual-like agents to drift in and out of the organisation, the reward for work being ownership. Intriguing stuff. A great essay on the topic is A Prehistory of DAOs (by @keikreutler).


How an organisation’s self-determination, ownership, and value-creating work become indivisible, held by the same people: the employees?

It turns out this same question is being asked at all scales.

So let’s assume the magnetic pole flip is in progress!

Or at least, let’s assume this: there is tectonic tension towards this corporate ownership inversion, even if as yet unrealised. So enabling tools will quickly find traction and unlock behaviour.

Answering questions like…

  • Simple paperwork to flip a company’s equity into an Employee Owned Trust, as with ustwo and Clearleft, and case studies of the same?
  • Software to enable worker-owned co-ops (as previously discussed).
  • Common patterns for meta-organisations such that (say) employee-owned design agencies can co-market and compete with global roll-ups?
  • Safety net/pension analogues for independent workers to buy into, and one-click personal incorporation, such that they have security while their employers can treat them with flexibility?
  • At a really mundane level: shared channels in Slack provide a way for teams in different orgs to collaborate. What other cross-org demilitarised zones might there be?

And so on.

If you were a VC you might invest in this, as a long term bet.

More News from this Feed See Full Web Site