The Five-Dollar-Uber Era of AI

A year ago was probably the best time to start. Right now is the second-best time because we can finally see the tide moving.

AI tools that would have sounded unreasonable for an ordinary person or neighborhood business are sitting behind twenty-dollar subscriptions. Some are free. One person can research, write, prototype, translate, analyze records and build software with abilities that recently belonged to a department.

The access is real. So is the bill hiding behind it.

We are still inside the cheap window

The largest AI companies are competing to become the tool we open first. They want developers building on their models, companies arranging their workflows around them, and ordinary people treating an AI assistant as the natural place to begin a question.

That competition keeps useful intelligence surprisingly affordable. Hardware improves. Models get smaller. Providers cut token prices. Open models create alternatives. Yesterday’s expensive capability eventually reaches a laptop or an ordinary subscription.

Yesterday’s frontier becomes consumer-grade only after the people controlling today’s frontier have moved on.

That does not make the present “generosity” imaginary. It makes it temporary leverage. We have access to tools that can extend what one person is capable of doing. The mistake would be assuming the terms around that access will remain this simple.

The price increase may not look like a price increase

The shift is already arriving through smaller decisions. A free tier gets advertisements. A flat subscription develops premium-request allowances. A preferred model disappears from a plan. Guaranteed capacity moves into an enterprise contract. An experimental service stops accepting new customers.

This is more likely than one giant industry-wide price hike. Basic AI can get cheaper while the useful, reliable, deeply connected part becomes something companies can meter.

Common people may never receive the frontier while it is still the frontier

We may get better AI every year and still fall further behind the people controlling the best AI. Large corporations and governments can reserve capacity, connect frontier systems to private information, and spread them throughout their operations before the rest of us have understood what changed.

By the time a capability reaches the ordinary consumer, the frontier may have moved again.

This is not a claim that ordinary people will be shut out of intelligence. The intelligence and capabilities will become generic over time. The class problem is timing. The groups with early access can use the frontier to accumulate money, information, distribution and operational advantage before yesterday’s frontier becomes our affordable product.

We can become dramatically more capable without moving any further beyond the working class. Capability is not the same as ownership

If a small group owns the models, chips, energy contracts, data centers and channels through which our work reaches the world, the tools can lift us while the economics hold us in place.

Consumerism gets easier when research becomes a flick of the wrist

Search used to require a small campaign. You recognized a need, formed a query, opened several pages, compared conflicting claims and decided whether to buy anything.

An assistant compresses that into one question: What should I get?

That convenience is useful. It also places enormous influence at the point where information becomes action. The system can decide which products appear, which sources matter, which tradeoffs receive a sentence, and how quickly curiosity becomes a purchase.

The companies do not only want to sell access to intelligence. They want to occupy the moment when an answer becomes a decision.

The mountain is moving

Those willing to understand the tools can already see that the current wave of corporate “generosity” will not last forever. It may not crash as one dramatic event. The wave can recede through advertisements, limits, queues, bundled services and products that become difficult to leave.

Meanwhile, the collective consciousness is lagging. Many companies are still deciding whether AI is a better search box. Others are already redesigning work around agents, private data and reserved computing capacity.

The mountain paths are becoming steeper toward the top. Behind us, the ground is shifting too. This is the moment to climb while access is broad, the tools are affordable, and the distance between ordinary capability and the frontier is still small enough to cross.

Build now, but keep what you build

The answer is not panic-buying every product with a waitlist. It is using this period to acquire knowledge and create systems that remain useful when the vendor changes the deal.

• Keep source information and customer records outside the AI provider.
• Save important instructions and prompts in files you control.
• Store the final business record separately from the chat.
• Test whether a second model can complete the same job.
• Maintain a manual path for consequential work.
• Measure value per completed task, not excitement per demonstration.
• Treat the model as a replaceable component rather than the company’s memory.

Use their wave of “generosity” to build something that remains yours after the wave crashes.

We may never own the frontier. But this may be the moment when we are close enough to take something lasting from it.