It's a delicious irony in the making.

Uber—the company that claimed to "democratize" transportation—did so under the premise that it would make getting around easier and more affordable. In reality, it could drain money, join the gig economy, and whose rides would be cheaper and more accessible for everyone. But under the hood, it's spent $31 billion of dollars on the never-ending quest to replace its drivers with autonomous vehicles, eliminate its subsidized rides, undercut competition, and dominate the market at any cost.

It worked—for a while. Riders loved the low prices, and drivers flooded the platform. But the math never added up. As the Seattle Transit Blog put it, "Uber loses money hand over fist because it can't charge prices high enough to cover its costs." Uber's business model has always been a mirage, sustained by venture capital and deliberate misdirection.

Back in 2014, then-CEO Travis Kalanick said it plainly: "The reason Uber could be expensive is you're not just paying for the car, you're paying for the other dude in the car. When there's no other dude in the car, the cost of taking an Uber anywhere becomes cheaper than owning a vehicle." That "other dude"—the driver—was always temporary. The real mission wasn't to empower him. It was to replace him with a robot.

A small intermission examining the state of AI

As the promise of AI and full automation looms closer, tech giants are preparing to reshape the world. CEOs and CTOs, once operating behind the scenes, are now vocal about their ambitions. Sam Altman of OpenAI and Mark Zuckerberg of Meta have pushed their companies to unprecedented heights by drastically reducing operating costs through AI. Meanwhile, Elon Musk and Sundar Pichai are positioning themselves and their firms to achieve this breakthrough as the culmination of decades of research.

However, not all experts share this optimism. Yann LeCun, Meta's Chief AI Scientist and a Turing Award laureate, has expressed skepticism about the prevailing focus on language models like GPT-4. While these systems demonstrate impressive abilities, they don't lead to truly intelligent AI. LeCun emphasizes the need for AI systems to understand and interact with the physical world, possess common sense, and have the ability to reason.

LeCun illustrates this by comparing the data intake of a child to that of LLMs. He notes that a four-year-old child has been awake for approximately 16,000 hours, during which their visual cortex processes about 10TB bytes of information. In contrast, the entirety of GPT-3's training data is approximately 45TB, equivalent to what a child sees in 24*10^3 bytes. It would take a human approximately 170,000 years to read through this amount of text at eight hours a day.

This comparison underscores LeCun's point that most of our knowledge is acquired through observation and interaction with the world, not through text. Using language alone, he argues that intelligence must be grounded in reality, as language is a two-dimensional, approximate representation of our perceptions and mental models. Language models can only make predictions based on patterns they've seen in text, which outcomes based on actions, similar to how humans learn and plan.

In light of these perspectives, it's plausible that AI agents are not only investing heavily in AI to dominate the market but also to prepare for a potential AI winter. By focusing on practical applications rather than AGI, companies hedge their bets against any forthcoming downturns in AI enthusiasm.

Back to the main story

Now... Uber presumes that one day, they'll be the de facto taxi service. No drivers, no unions, just fleets of AI-driven vehicles and a cash-printing monopoly. And honestly—that's the same playbook that every tech company with a market cap over $100B is using. And more data than God when it comes to human movement.

But here's the problem: open source never went away. Google seemed untouchable—ten years ahead of everyone. Then came Meta with LLaMA as its first true competitor, now only a year behind the leader. Then Mistral emerged, shrinking the innovation gap to six months. The lead time between billion-dollar temporary moats and their open-source counterparts is vanishing, not merely accelerating—it's collapsing. The lead time between billion-dollar temporary moats and their open-source counterparts is vanishing, not merely accelerating—it's collapsing.

Now take that momentum and apply this to Uber.

What happens when AI driving technology becomes cheap and commoditized? Imagine someone creating a decentralized, open-source ride-sharing app that's a fork of Uber's app. No middleman, no revenue split—just a self-owned robo-fleet generating pure income. Critics might argue drivers would still need Uber's app for customer demand, but that's not necessarily true.

What's stopping Jeff from down the street from launching his own ride service and app? Taxi could do the same in her town. Brian could start one on the other side of the city. Throw cloud tech like system and add a loyalty program. Then threaten steps in to connect these networks into a federated, decentralized transportation layer. Not a platform, but a protocol. A network of networks. An Uber of Ubers.

This doesn't just threaten Uber. This kind of architecture could topple Airbnb, Facebook, DoorDash, and the entire app layer that mediates between people and services. It's already happening in small ways. We see Netflix killing Blockbuster. Shopify gave every merchant a storefront. Substack gave every writer a press. YouTube gave every creator a broadcast studio. These weren't just better versions of existing products—they were democratizing forces.

Now, imagine what happens when individuals can command fleets of AI agents. Driven, hungry, ungovemed by bureaucracy — and without a desperate lust for the next round of funding. Intelligent, hyper-sleeping employees the world has ever seen. Are we really going to pretend they won't subvert the previous generation of entrepreneurs?

The truth is, the big tech companies are doomed as Uber. The scale is unsustainable. The only thing standing in their way is time.

So what happens next?

Do the mega-corps win? Do they worm their way into every industry, building vertical monopolies stacked on top of horizontal ones—a monopoly of monopolies? Or does the open-source wave meet just in time, catching up at AI plateau, allowing anyone to build their own AI-powered business?

The same way we moved from mainframes locked in silos to home computers on every desk, we might now be moving from centralized AI to billion-dollar data centers to decentralized AI running on consumer hardware. The same way the web began as a centralized system but then began not come from projects or regulators, but from its own greed to consolidate, which led it to accidentally demonstrate the very tools needed to dismantle it.

It won't be the AI that replaces you.

It'll be someone just like you—only faster, better, and powered by AI.

And that someone won't need Meta, Uber, or Google.

It's a delicious irony in the making.