The Efficiency Economy

The US economy in 2026 is leaning almost entirely on the abstract ‘promise’ of AI technology. With roughly five monopolies driving the stock markets with AI centric valuations, its reasonable to state that global economic stability is bound by the performance of these corporations, and as such, the continued perception that the LLM technologies these companies are investing in will deliver something of tangible value to the broader economy, i.e all of us who have a relationship with money (jobs, spending, investing etc). 

The reality, however, is that as I write this in early 2026, there is no tangible value in AI for the average participant in the ‘real’ (non-stock market) economy. Let’s break down why this is. 

The current value proposition of the handful of companies developing and spreading LLM technologies to the public is one of efficiency. You see this buzzword on display in nearly any company, large or small, promoting AI agents and tools that make something they want you to do ‘more efficient’, ‘optimized’, ‘disrupted’…’revolutionary’.

The efficiency label is just another way of saying replication. LLM tool output is efficient because they replicate. They replicate human labor in the form of accumulated knowledge. They are massive data aggregators that leverage 60 years of research and advancements in computing to distill the entirety of human knowledge as it has been documented and archived in the Internet over the past 30 years. I will qualify the term efficieny, however, since the energy consumption and enviromental impact of massive data centers that power LLMs is obscene, and in this way, they are anything but 'efficient'. 

It's worth pausing here for a moment because although this piece is an expression of apprehension about the ways in which unregulated AI is directing the economy, it’s power is amazing and it does have potential for real good in certain domains. Organizations and companies using LLMs to advance scientific research are one example of a potential innovative good coming from this technology, and this still seems to be an under-reported and under-represented sector of the larger AI conversation. 

But this conversation must start with regulation, and in any case, I digress. 

The idea of replication without innovation is something worth exploring because it centers on the concept of perceived value. Perceived value is a marketing principle that is used to influence a customer’s opinion of a product, and is leveraged to guide pricing and decision making. This is foundational marketing to get a product to stand out from the competition by not just emphasizing what it functionally does, but by emotionally appealing to a customer to get them to perceive the product in a certain way. Large brands do this really well by convincing us to buy an image of ourselves or a lifestyle, not just some object. 

The companies that unleashed AI on the world four years ago have been working tirelessly to shape public perception of the product as valuable to people’s everyday lives. The inherit problem is that the perceived value never evolves beyond perception into actual economic value. Any value someone gets from querying information or completing some task using an LLM is recycled value. It is not new. It is value repackaged and consolidated into a new format, often sprinkled with invented (hallucinated) information. It may be fast and solve an immediate problem, but it is fundamentally nothing you couldn’t do on your own with the same information from another non-LLM source. It is an illusion of a tangible economic outcome. 

The average person using an LLM chatbot at work to deliver a task for a business is not gaining more economically from someone who isn’t. The company may be collecting the difference of any financial gains but the person doing two people’s job with the aide of an LLM isn’t making twice as much in wages. Just as if you were taking on more responsibilities at work with no additional pay, the fundamental imbalance in productivity / reward persists with AI. Therefore, any perceived value remains speculative as circular investment deals persistently juice the markets, further inflating the values of a few monopolies. Beyond shareholders, executives or other higher order employees of these corporations, no one’s life is functionally improved in any meaningful, measurable economic metric from AI technologies.

The social division, misinformation, and abuse perpetuated by LLMs is well documented already, so one remaining philosophical concern is: what does the individual gain or stand to gain from a world dominated and governed by AI tools? As we’ve established, the concept of ‘efficiency’ is driving the AI value proposition, but efficiency doesn’t usually create improved economic circumstances for workers.

It also doesn’t create innovation. 

Doing something faster isn’t by itself innovative. The Internet was innovative because it became an infrastructure for entirely new forms of communication and transaction - email, e-commerce, digital banking, etc - efficient because they were innovative. None of these eliminated their analog equivalents. People still send and receive physical mail, shop in physical stores, and go to physical banks. The digital versions modeled the analog versions, but innovated a reach and accessibility that improved people’s lives and created conditions for entirely new business domains within each system. The core difference lies in an ethos of competition and divergence vs one of monoculture. The argument isn’t that the modern Internet is not trending towards monoculture (it is); rather, that the early Internet facilitated a marketplace of ideas that functionally improved economic wellbeing of a large number people over the last couple decades. Sure, monopolies emerged, oligarchs were minted, but there was also a widespread economic democratization of opportunity as products and services were created within the digital space. 

AI, however, is pure monoculture. It steals and redistributes existing ideas and work and suffocates diversity and originality by appropriating everything to itself. To the extent in which it creates any new business or market opportunities, they are replicas of the same idea and core conceit: obsoletion through automation.

Bot ‘assistants’ abound to ‘help’ you with your order return, reviewing resumes, writing a text snippet for your email newsletter, or dozens of other trivialities that in another era would be accomplished within a few minutes of sitting undistracted and just thinking. This is not innovation. 

Anyone who has spent time interacting with these agents senses that these tools are awkward and ineffective at solving real problems. They all eventually require a human mind at some point in the transaction to actually accomplish anything, but corporate executives - hungry to keep pace with the markets - have mandated these tools be integrated into every aspect of business irrespective of their utility or user sentiment against them. AI is a hostile takeover of the human intellect and social protocol.  

AI has been trained on everything anyone has ever created on the Internet and purports to regurgitate it on demand as if it were original thought or unique content. The remix is the novelty, a toy-like distraction to keep us oblivious and passive to an increasingly anti-democratic techno-feudalist system where a few billionaires operating a dozen or so corporations own and surveil every aspect of our lives. Corporations are comfortable doing this, of course, because we have given them permission. 

Everything we’ve created on the Internet for the past twenty years has been appropriated to the plutocratic class through terms of use agreements. Social media at least had the decency of a community pretext to get you to part with your knowledge, your creativity, and your personal data. You also knew (or should have known) what you were giving up by consenting. AI as a force has none of this. It is pure theft. It was silently stealing from us for years before it was released to the public as a chatbot.

The consent to AI is merely existing in the current economy, which is not consent. Its only mandate is to take; its only deliverable is the vagueness of operational efficiency. If there is to be any meaningful innovation within AI, it must be against our own collective, rapidly approaching displacement.