Technology is the Problem

An alternative cause for the Great Stagnation: the cargo cult company

In 2011, Tyler Cowen argued that the developed world economies were in a Great Stagnation. It had started in the late 1970s because all of the low hanging productivity fruits—machines and factories powered by fossil fuels and electricity—had been exhausted. 

Technological improvements were happening at the margin.  This same sentiment is captured in Founders Fund’s famous saying that “we wanted flying cars, instead we got 140 characters.”  

And this tweet captures the perfect banality of most technology today.  We entered a SaaS hole of darkness.

My colleague Ted Mabrey wrote that “Software companies are shipping more, institutions are consuming more and the result of all of that software consumption is productivity growth below the long term trend.”  

Satya Nadella at Davos in January 2024 said; “ ...inflation adjusted, there is no economic growth in the world, I would say and that's a pretty disappointing state. In fact, the developed world may have negative economic growth...PCs were the last time, when actual, economic growth came about, right? So, the last time it showed up in productivity stats were when PCs became ubiquitous.” Satya goes on to argue that AI has the potential to restore productivity growth.

While AI is very powerful, this argument is wrong.  AI alone is insufficient because it fails to diagnose the underlying systemic drivers of the productivity dysfunction, let alone redress them: implementation for advantage.

I propose an alternative cause largely born out of 18 years of experience at the coalface of delivering productivity to governments and Fortune 500 institutions. The Great Stagnation is caused by technology and not despite it. Software in particular is suited to layers of abstraction turned obfuscation in a way hardware is not. AI only accelerates this trend. We are stuck in the era of the cargo cult company:

During WWII, the U.S. used islands in the Southwest Pacific as military bases. The indigenous Melanesians observed with awe as big metal birds carrying all sorts of supplies and cargo magically materialized whenever the soldiers were marching and signaling towards the sky. When the war ended, so did the appearance of the magic cargo. Local chiefs tried to summon the cargo by engaging in the same rituals they believed caused the cargo to appear in the first place. They cleared the forest to look like a landing strip and made wooden replicas of airplanes and control towers. The Americans and their Coca-Cola did not reappear. 

Today, “cargo cult programming” is well known among engineers to refer to ritualistic code practices divorced from cause and effect. It is symptomatic of an engineer who doesn’t understand the underlying problem and tries to pattern match to get at the solution. The analogy is not often extended to describe cargo cult management, but this is an even more pernicious problem. In cargo cult management, a company’s leadership uses the employment of a technology (frequently software) and the achievement of financial metrics as a proxy for actually solving the problem and obtaining ground truth. 

Cargo Cult Companies Rely on Abstractions Turned Obfuscations

There is an old saying in Silicon Valley, “There is the first 80% and the second 80%.” While it is really hard to create new technologies, it is also really hard to implement them for any measurable advantage.  This has always been true: The steam engine didn’t matter until it was put into a ship and locomotive; the Wright brothers’ flight didn’t matter until it moved people; electricity needed to be delivered to the home; and telephony didn’t matter until there was a connection. However, the connection between invention and implementation used to be more obvious. 

In the 1970s, complexity in technology reached a point that it wasn’t possible to move forward without abstractions. I was taught the beauty of abstractions as an electrical and computer engineer. Doped silicon where you are solving physics equations suddenly become transistors where you are dealing with 0s and 1s suddenly become chips where you are issuing instructions, then computers where you are writing code, and then user interfaces where you are clicking. It was a great way to deal with overwhelming complexity. The abstractions made it all manageable. 

But now the abstractions are getting in the way.

We have converged on a world where the people who do the first 80% have no competency or interest in doing the second 80%. The SaaS industry is a frequent target of the Great Stagnation because it is chasing high margins rather than high value outcomes - which, to be fair, is a rational response to a punitive capital landscape that penalizes companies for deviating from a specific set of metrics. And it’s not just SaaS. The Defense Industrial Base underwent a financialization in the 1990s where the focus shifted to dividends and buybacks rather than innovation. I could go on. Lacking the incentives to allocate resources towards implementation, companies instead create an abstraction to define the “right” way to do something. 

But when the abstraction isn’t tethered to empirical results or anything measurable and falsifiable, the abstraction is just an obfuscation, a cargo cult ritual. You can’t prosecute the AI revolution without piercing the abstractions substantially to redesign the application, its code, and the chip (CPU→GPU) it runs on. SpaceX succeeded not because it invented a fundamentally new rocket or law of physics but because it didn’t rely on crusty abstractions that obfuscated what was actually physically possible. 

When I look at the commercial world at large, despite years of investment in technology, it is completely broken. The cargo cult company’s latest billion dollar enterprise resource planning (ERP) implementation obfuscates the fact that employees on the factory floor are still running production in Excel. Similarly, the strongest indictment of the software-industrial-complex is COVID. The only technology that companies could point to as relevant to dealing with the disruptions of COVID to their supply chains and operations were videoconferencing software like Teams and Zoom. What about the 100s of billions of dollars invested in the modern data stack, cloud, ERP,  data lakes, data warehouses, and CRMs? 

The problem isn’t that enterprise software doesn’t work - it is that you didn’t make it work for you. The only measure that counts, hidden behind all those layers of abstraction, is the final outcome - decision advantage.  This is often viewed as an implementation detail (the second 80%), and implementation is often outsourced to experts (COVID should have also taught us all something about “experts“). The industry of consultants who themselves have never run businesses but tell cargo cult companies how to execute implementation further obfuscates the lack of productivity gains.

Palantir was ridiculed endlessly for our Forward Deployed Engineering model. Investors would impugn Palantir as a services company simply because we didn’t believe that throwing our software over the wall for consultants to implement was the correct thing for our customers.  In fairness, many of our buyers hated this because they subscribed to a world view where engaging in the correct rituals (e.g., buying expensive ERP) was the only thing that should matter. Like Department of Defense program management, the commercial world often believes implementation is best executed when managed to cost, schedule, and performance by fungible units of “independent” consultants. 

Palantir never charged for our implementation services because we wanted to internalize the cost and the pain.  The only way to make the software implementation better was to commit ourselves to solving the whole problem.  SAP, Salesforce, Microsoft, Snowflake, and so many others are responding to what the market wants: high margin software revenue. They rely on a scaled ecosystem of integrators to do services, and that’s exactly why we are consuming more technology and none of it matters. 

John Boyd would reframe this as being about the OODA loop. Observe, Orient, Decide, Act.  The technology that matters accelerates your OODA loop by helping you make decisions faster than the competition.  And that compounds.  Everything else is rearranging deck chairs. At best you are smarter, but you aren’t better.  A lot has been written about the OODA loop already, but I think the best summary comes from JRock in his (brutally crass) anthem: Win.  Or more eloquently by former Chief of Staff of the Army McConville: Winning matters.

To say the quiet part out loud: the OODA loop is all about the human.  Technology serves humans.  Instead, we have cargo cult companies such that the mere acquisition of architecturally-ordained technology is perceived to be the solution. The cargo cultist treat the front line users as godless heathens standing in the way of adopting innovation.  

More profoundly, even the purveyors in the software-industrial-complex really believe their technology is working (Satya proves to be the exception with a more clear-eyed understanding of what is actually going on). They too have subscribed to the cargo cult because they don’t go to the factory floor; they don’t own the burden of implementing their technology; and they aren’t accountable to delivering productivity. Management at the company won’t pierce the veil of abstraction to realize they fell for the Wizard of Oz. 

How to Leave the Cargo Cult

We overcome these problems by recognizing that great leadership is more important, not less important, with greater technology.  The more sophisticated and complex our technology become, the more important the greatest leader becomes.  Intel’s road to ruin was when the CFO became the CEO and managed the business by the numbers and abstractions. Former Intel CTO and current CEO Pat Gelsinger might correct the course.  In the meantime, technical founder and NVIDIA CEO Jensen Huang rewrote the stack around accelerated computing.

In this way, technology has largely had the opposite effect from all the effects anticipated.  The world thought technology would be a great leveling force, but it has led to more winner-take-most, power law outcomes. Technology was supposed to enable the median human to do so much more, but it has actually made the very best much more valuable. The world thought technology would make geography less important, but it has made it only more important by concentrating the relevance, wealth, and share of innovations from a handful of places.  Silicon Valley is more relevant, while Europe lags further behind in Tech. China still has zero global enterprise software companies. Tim Cook says that Apple makes things in China because they are the best, not because they are the cheapest — a consequence of a flawed industrial policy in the West.

Steve Jobs called the Macintosh a bicycle for the mind. Let’s abandon cargo cult “solutions.”. Recognize there is no process, and it will be painful. Technology only works when it is an extension of the human mind and the humans who mind it. Exclusively work backwards from decision advantage.