The Biggest Problem Is Solutions.

The biggest problem in the world is solutions. There’s an app for everything, a vendor for everything, so why isn’t everything perfect yet?

The answer is that solutions have costs. First among these is that you don’t have time to use all of these apps: you only have 24 hours in a day, but every year, there is more and more technology for you to use. That is, there are usage costs.

The most overlooked economist of the 20th century, Ronald Coase, focused on friction. He noticed that friction between supply and demand — discovery costs, transaction costs, integration costs, switching costs, delegation costs, and usage costs — defined the shape of the economy. Like a physical law, these frictions explain why one company doesn’t do all things, why there’s an app for everything, and why everything still isn’t perfect yet.

What results from usage costs is a ruthless competition to out-use technology, that forces everyone to work harder and harder to extract the most utility from these tools. Technology was meant to liberate us. But everywhere we are in chains. Instead of being the masters, we have turned into the slaves.

The Greeks had a word, dynamis: the potential that exists within things. Every year, the dynamis of technology increases, but our usage only captures a small fraction of that expanding potential.

In the broad brushstroke history of Sapiens, Harari dedicates an entire chapter to “Memory Overload” to make a simple but profound point: after the agricultural revolution, writing was less of a breakthrough than bureaucracy. Writing emerged in many places, but only those societies which built bureaucracies capable of efficiently organizing written data were able to harness the dynamis of writing, “opening the way for the appearance of cities, kingdoms and empires.”

In the Chinese sci-fi series The Three-Body Problem, a hostile and advanced alien race prepares to invade earth in two hundred years’ time. Wary of human progress, it sends miniature probes in advance of its fleet to block any progress in fundamental physics, so that humanity can’t defend itself when they finally arrive. But even with a scientific blockade preventing progress in physics research, technology continues to advance, making remarkable progress.

How is this possible? Even though technology is limited by science, progress continues because, so far, technology has only captured a small fraction of the dynamis of science — it is already hundreds of years behind, so there’s room to catch up.

In the same way, even if technological progress stopped today, and there were no new apps, it would take us, say, two hundred years before we caught up to the usage potential of the personal computer, the internet, and cloud-based software applications.

Steve Jobs imagined the personal computer as a bicycle for the mind. But although more and more advanced personal computers are built every year, why haven’t we realized Steve Jobs’ vision for human renaissance; why isn’t the average knowledge worker, say, twice or three times as productive as Leonardo da Vinci was, half a millennia ago, without these tools?

The reason is that solutions have usage costs. Personal computers, the internet, and cloud-based software applications exist, but we’re only able to take advantage of a fraction of their power and potential, their dynamis.

The dynamis of Google is to provide access to all information on the internet, but usage costs limit us to the top three results. The dynamis of Google Drive is to organize your digital brain, but the usage costs limit us to scattered files and folders. The dynamis of Facebook is to intimately connect you with your friends, but the usage costs limit you to responding to only the top messages and notifications. The dynamis of LinkedIn is to provide access to the entire economy, but the usage costs limit it to a glorified resumé.

So why haven’t we solved this mysterious problem, the problem of solutions? Is it even solvable? Yes, it is a solvable, practical, problem. And it should be solved in our lifetime. But blindspots and biases have led us down blind alleys.

While the venture capital industry has indulged in endless self-loathing for its gender biases for men over women, it has ignored its institutional bias for products over services. Every aspiring unicorn builds its own walled-garden software platform with a restricted API, creating a tragedy of the commons in terms of usage costs. The next Salesforce can’t build on top of Salesforce, or make Salesforce ten times more useful — without limiting its destiny by violating Machiavelli’s doctrine to “not build upon the power base of another.” PayPal made this mistake by building on top of eBay, which ultimately forced Peter Thiel’s hand in acquisition, although he knew that ultimately PayPal’s value would eclipse eBay’s.

Venture investors have an equally severe bias for vertical over horizontal solutions. Ricardo’s specialization has led to hyper-specialization, but has failed to progress to meta-specialization — the dialectic is stalled. Applications are increasingly narrow, and returns driven by acquisition instead of IPO have made it harder for large funds to compete with proliferating micro funds.

VCs like X-for-Y pitches (Uber for laundry!), they don’t like Y-for-all-Xs pitches (Youtube: videos for the entire internet!) anymore — times have changed; that was when the internet was still young. Ironically, in the young blockchain bubble, there was perhaps too much horizontal, and not enough vertical. But in the relatively mature industry of cloud-based software applications, venture investors are suspicious of horizontal breakthroughs that aren’t driven by some fundamental technological breakthrough.

While investors are aware of cycles of bundling and unbundling, after a decade of unbundling, nobody is preparing for the inevitable bundling. The smartphone cycle bundled many devices into one device, the app cycle unbundled those devices into different apps, and now the AI digital assistant cycle is moving towards a single bot that can do everything: ultimate power and simplicity at the front end, ultimate complexity at the back end.

What is the natural relationship between General Artificial Intelligence and assistants? In the 20th century economy, assistants solved for Coasian frictions, empowering executives to be more creative and strategic. They used the bureaucratic machinery, so you didn’t have to. But only VP-level executives and up can afford assistants. And even the best assistant in the world can’t work more than 60 hours a week.

So if you need more leverage, your assistant needs assistants — and that’s exactly what heads of state and executives of Fortune 100 companies did, and do. Another drawback is that assistants are generalists, they don’t have time to specialize in everything that they do; if they did, they would be meta-specialists. When a good assistant runs into the limits of their usage time or specialization abilities, they connect you with more specialized resources: both employees inside and vendors outside your company — but they do this inefficiently, operating on intuition rather than advanced systems built on organized information.

What AI promises is an infinite delegation resource — a meta-specialist, a single touch point for unlimited specialization, an frictionless actor — which can be used by everyone in the organization: solving all usage costs, except for the cost of delegating to the AI. The ultimate bicycle for the mind.

Following Coasian logic, in this futuristic post-AI economy, companies would get larger and smaller: hyper-efficiency would enable most companies to become smaller, and a few companies to become larger, shrinking or growing by one or two orders of magnitude. For example, the largest employer presently has 2M employees, but in the future perhaps there will be a company with 20M or even 200M employees. And there would be an even larger ocean of freelancers in the economy than there is today, frictionlessly contracting for the lake of small businesses and the island of enterprises. When the usage cost variable is altered, I imagine extraordinary change.

Despite all the hype about General Artificial Intelligence, VCs are smart enough to know that startups can’t compete with tech giants in heavy R&D — so the industry of ambition shies away from its own greatest ambition. VCs are investing in narrow AI applications, but again, they don’t solve the overall usage costs problem — they just add more complexity and power to the system, for users to compete to extract.

And tech giants are smart enough to indulge the hype, even though a General AI assistant capable of doing meaningful knowledge work is not actually on the horizon. Even though it is a moonshot, they keep investing in it because whoever owns the assistant relationship with the consumer is a gatekeeper for all other applications and vendors: a monopolist’s dream.

So if General AI, the pure software approach to solving the usage costs problem, is a fanciful science project pending 22nd century completion, two approaches remain for the 21st century.

Tech-Enabled Digital Assistant Services like Fin use technology to coordinate humans executing knowledge work. But Fin shut down to pursue its software business, reinforcing the institutional bias against services.

In theory, this approach makes sense. If you want a service to use your apps for you, to take over all repetitive digital work, you could employ an army of humans to do it for you. But you’d need the technology to coordinate them and secure their access. If you could overcome these obstacles, the arbitrage exists: you trade your usage costs for their usage costs — problem solved. But even with cheap labor, without automation to drive down usage costs over time, this is still too cost prohibitive for consumers, although not for enterprise.

Robotic Process Automation is the other approach. RPA companies work with big enterprises to carefully define high volume processes, then automate them. For example, a large bank may use RPA to automate 20% of its accounts receivable staff. The approach is slow, expensive, limited to enterprise buyers with defined processes running at scale.

Three horizontal, meta-specialization approaches — General AI, Tech-Enabled Digital Assistant Services, and Robotic Process Automation — are underway to solve the biggest problem in the world: solutions. All of them are flawed. Perhaps the right combination of them will succeed. Can Theseus find his way through the labyrinth?

If you asked most people, after ranting about Trump, they would say that the biggest problem in the world is General Artificial Intelligence replacing humans with robots, and that the solution to this problem is Universal Basic Income. Bah! The irony! Software won’t replace us with robots anytime soon, but software has already turned us into robots! What we need is a solution to free us from solutions, software (or software-enabled services) to free us from software, so that we can be humans again.

Am I here to take your job,
Or are you here to give me mine,
Or am I here to teach you yours, human?
— The Robot’s Koan

In The Question Concerning Technology, Heidegger suggests that the way for humans to enter into a freeing relationship with technology, is for humanity to unlock the dynamis of technology, so that technology can unlock the dynamis of humanity. The message is: fear not! That is, don’t fear the solution, fear the problem. The problem is that we’re slaves to technology: there’s a solution for everything, except for this Sisyphean usage curse. We’re wasting our lives, squandering our creative potential, doing work that could be done for us. So, solve the usage costs problem! Build a meta-specialist. Build an infinite delegation resource. Build a single touch point. Build a frictionless actor. “Forth, and fear no darkness!”

What would the world be like, if solutions had no usage costs? If the economy was free of Coasian friction? Technology is at its best when it is invisible. You wouldn’t even notice all of the products and services empowering you to be creative, strategic and human. You would never need to do work somebody else could do for you. You wouldn’t need to be organized. Returning to Jobs’ humanistic vision, in The Once and Future Renaissance — “Free at last! Free at last!” — every individual has a bicycle for the mind. And then, who are we, and what do we do; are we capable of genius, and is spirit moving? These are questions for Nietzsche’s supermen of the future: once we have delegated our work, the real work will finally begin.