July 07, 2026
Everyone, including myself, has tried to invent some perspective or take on the current “AI bubble”. You could compare it to the recent crypto bubble where everyone seemed on board with this “technology of the future” but see it falls short in how crypto never broke out of its own “world”. DeFi only works in the world of crypto where AI is already used in different industries. You could compare it to the dot com bubble where any company with a dot com domain was IPO’ing at insane valuations. But, then see it falls short in how the internet was the first time people saw the widespread technological change happen in real time; consumers are aware of previous bubbles and not looking to have the AI hype be a exact repeat of the past. Any past “hype cycle” has a strong caveat which makes this AI phenomenon impossible to accurately compare without some “but really X was different” disclaimer.
However, a take I heard mid-conversation from someone who heard it from someone else stuck with me. What we’re seeing in the tech industry (especially with regard to programming roles thanks to coding agents) is similar to what happened to formerly Communist countries when they opened their borders after the collapse of the Soviet Union. It’s not a clean direct metaphor so, to deliver the point, let’s imagine we’re someone who works on some assembly line in a factory for a home-grown product.
The product you work on isn’t the greatest quality but it’s precious, it’s marked with the direction and instructions of the nation you’re producing value for. Your friends working in non-State-backed markets seem silly to you because they must cope in some way when their trade isn’t performing well. You’re smart for contributing to something the very nation you’re a part of says is worth something. Then, one day, that security and dependency disappears, democracy is trendy, and your neighbors can purchase from abroad rather than locally produced. While you protest consumer choice, you can’t help but notice your industry go belly up practically overnight.
Suppose, one day, North Korea stops being an authoritarian state and everyone who used to only buy from local manufacturers can buy from a neighboring Asian country. Assuming we’re past the “calibrating purchasing power with the rest of the world” phase, the majority of people would be expected to buy from abroad if it’s better or cheaper. Call it free market capitalism or competition, but the change in consumer behavior would be clear. Every SWE sitting with guilt or uncertainty over their future career is experiencing something similar to factory workers in a formerly shut-in nation that’s now opened up its borders.
If your immediate reaction is similar to when tech workers ask for a union, pity for techies is not the point of this post. Regardless of the average SWE salary to global poverty baselines, coding agents being able to replace technical talent is revealing of underlying problems similar to how COVID didn’t worsen institutions but made their problems clearer.
Back in the day, technical talent was as non-fungible as brand identities. If you were doing a venture-backed startup, you’d accumulate engineers to bring to life the vision in your pitch deck. It was a required dependency since, otherwise, who’d make the app? Nowadays, you can vibe the entire MVP with nothing more than a prompt or markdown file if you care that much. People would establish their careers on being specifically familiar with a certain programming language or framework (remember when bootcamps were shipping React or Angular developers?) but now that knowledge is only relevant insofar as it’s not immediately obvious to the models.
Whenever you were faced with a problem in the past, there was an actual tradeoff between finding something off the shelf or making it yourself (the classic “build vs buy” argument). Nowadays, people are able to make their own in-house solutions as soon as something off the shelf gives even the slightest inconvenience. While specific needs or “enterprise” use cases may require the existence of focused companies on certain solutions (ie Salesforce or Datadog), open source work or lower token costs help in the direction of people being able to build their own solutions. There was a time when it seemed crazy to leave bespoke hand-crafted books for the printing press. There was a time when it seemed crazy to leave fine cultery for plastic utensils. There was a time when it seemed crazy to leave intentionally made things for mass produced items. There was a time when it seemed crazy to leave physical media for the internet. The same will happen to software and how it gets engineered.
Other theories of the AI bubble generally describe how it will behave economically today and around the peak of the bubble, but they don’t always describe how the world will change afterwards. With the industrialization of software, seeing how consumers responded to the industrialization of mass produced goods, it paints a narrative for how generative UIs shift from being toys to real applications. An ordinary consumer looking for an item will go to Walmart or Amazon first before they look for a small local business that hand-crafts that said item. A UI that provides to users just what they need informationally makes a lot more sense than an app that was engineered for a world of capturing attention.
While apps are technically playing the same game as the “attention economy”, the goal is slightly different. Before, you wanted consumers or users to be viewing your displays to get the value you provided. Now, you want consumers or users to be getting the value you provide, regardless of the display[s] you may provide out of the box. MCP turns your buttons and forms into text questions and answers. Generative UIs turn your pages into partially renderable views for users to get no more than they’re interested in. You don’t care if a person pays you after messaging a chatbot or after clicking through a series of forms so long as your service is ultimately provided.
Some things do not change, of course. An app that’s trendy or popular will play the same today as it did in the past with word of mouth. Programs do ultimately depend on the value they’re providing and not just their aesthetic value. But, funnily enough, the vibe coders got it right with being able to view software as nothing more than an incidental problem for an AI to deal with, not people. People making toys don’t care about the exact chemicals used with the plastics, they just want a plastic toy in the end. Likewise, people solving problems with software will critique much less whether an app was developed in some language or another. I think people can care or not care what language an app was developed in. We may find a world where people care less about technical implementation details or a world where people actually care about programming languages like the environmental impact of certain culinary categories.