Everyone can do everything with AI

With AI, everyone can build everything.

With AI, anyone can do anyone else’s job. Everyone can build a website or an app. They can even diagnose themselves replacing doctors. There are no limits anymore.

SaaS companies will be dead soon, as everyone can copy their product.

I see lot of people vibe coding projects on Reddit, and most of them openly say that they are not developers, but neither product owners nor marketer or experts of the domain. On X it’s even critical: scrolling it for a couple minutes creates the FOMO: “how is that everybody is building something useful while I’m losing my time on X? Why is everybody successful and capable of using AI while I am not?”

I always see these posts through Reddit and X. It’s becoming detrimental for mental sanity, self-confidence.

Is it really like this? Did AI give the power to everyone to build almost anything?

I don’t think so.

This narrative reminds me of the internet. The entire human knowledge is on internet, and no more than 20 years ago many people believed that universal access to information would make everyone more informed and capable of studying any subject. All the content was online, free, funded by advertising, and ready to be consumed. Existing limitations to resource access seemed to be demolished.

However, we all know that things didn’t go this way. Disinformation, internet platformisation, accessibility issues and environmental factors like instructions and personal curiosity slowed down the mass educational process. We confused access to knowledge with acquisition of knowledge. And I think AI will go in the same direction.

I’m a software engineer and I know how to build software. I spent years in writing code, drafting architectures, and everything else involved in the process of building digital products. According to the narrative however, just like anyone is building software, I should be able to do the job of almost anyone else in almost any field. For example, even though I know little to nothing about economic/financial domain (despite having some exposure to the field in high school), I should be able to work with or replace anyone in the job, who’s studied and worked for years in the subject. After all, LLMs are there ready to answer any of my questions, right? Right. But: 1) it takes time; 2) I don’t know what questions to ask; 3) I do not know what matters most; 4) I do not know what the best practices are; 5) I don’t know the domain vocabulary, team processes and methodologies; 6) I do not have a well working and tested method. And finally, 7) I have an entire knowledge and a mental process to build, which might require years.

The point is that although the tool (internet back then, AI now) has no intrinsic limits in terms of knowledge diffusion, people do. Society has limits too. Human beings have limitations in terms of capacity, time, energy, reasoning, commitment, passions and preferences. These limits can be eased by frontier technology, but cannot be smashed.

If you’re unsure about it, dive into someone else’s field: design jewellery, design houses, build the easiest SaaS you use and try to replace it for your own use. Your MVP might work, but what about design, UI, marketing, accessibility, feature prioritisation? You’ll discover that with AI you can do more and better things. Your skills are amplified, especially if you have a solid foundation in reasoning and thinking. However, they’ll stick to being good enough, which is now dramatically easier than ever to achieve. To make something really good, though, you still need domain expertise and experience.

The good enough also raises other concerns. AI helps increase the overall average quality of individuals’ and organizations’ outcomes, making the world even more competitive. Those who don’t or can’t use AI tools may never be able to close the gap with those who do. Those who use AI tools will see their outcomes align with others who use AI tools, creating convergence. In fact, the difference in outcomes will shrink, and only a few people, especially domain experts, will surpass the good enough stage.

For instance, I’ve worked with backend systems since forever, but I can’t create websites - not even the most basic ones. Thanks to AI, however, I can now build websites in minutes (e.g. Relego landing page). That said, this website will be similar to thousands of others built with AI (especially in terms of text content). The average outcome improves, but the differences shrink. This is so true that people on the Reddit blame vibe coders as they can distinguish a handmade website and one built using AI in seconds. Yet, these websites are years behind of what a professional team of programmers and designers could produce, especially if they use AI too. Moreover, these websites may lack branding, SEO, conversion rate and all the details that really make a difference. A lawyer who creates a website using AI might not even know what SEO actually is, and therefore would not ask the AI the right questions – hitting the point 2 above.

In conclusion, as more people become more capable of producing good enough results, the competitive advantage shifts from execution to judgement and curiosity. You have been required to be able to connect different domains to your primary domain, to expand your capabilities and blast radius without delving deeply into other domains.

If you stop for a second and think about it, connecting the foundations of different fields is the main limitation of the human intellect. Throughout history, the number of people in history capable of moving deeply across disciplines (eclectic) can be counted on no more than two hands.

In short, you have been required to be better than others at being generalist.

This may seem counterintuitive given what has been said so far, but characteristics such as study, approach, method and general culture can provide the foundations to guide the AI and help you to expand into other areas. This requires extra effort, which society actually requires. Without these characteristics, AI alone cannot you.