blog

Introduction

There is absolutely no coming back , that is in the technology field, when it comes to AI and how it’s being used day to day. The technology is mature enough to be able to complete most tasks a junior will do, and can even reason beyond what we would normally try even as seniors, and none of this is new information to you. At the surprise of no one, that creates supply for a new wave of applications and programs, since everyone and their cousin can vibecode. I personnally know a farmer that uses Claude to create python automations for his excel sheets, and experiments with esp32’s since its cheaper for him to do so than buying commercial products. From the perspective of the farmer, AI is wonderful, as it allows him to do extra work he couldnt originally, tasks that require programming or product design knowledge, to be able to maintain his crops for cheaper. And the thing he said, title of the article, is what struck with me : its Good Enough. Yes he can now program with more ease then ever, but he recognizes that he needs to learn more abuot programming to actually make a decent product that helps him. That begs the question : How can we use a technology that’s good enough to innovate beyond what it can do? and at what point should we use AI to keep our human expertise?

The fools gold

The tale is as old as time. good artists imitate, great artists copy, like Steve Jobs famously said and gave credit to Picasso for the quote. I the pre-AI days, you would often see people creating Saas solutions for note-taking, coud storage solutions, and more services that already had open-source alternative, because these builders relied on the conception efforts of the previous open-source projects to build their own. But with the era or AI agents now able to work on the entire tech stack, the field is drastically different nowadays.

Previous builders usually had strenghts here and there in tech, which allowed them to invest time to innovate on features and have a good comprehension of the project and the codebase. Whenever they hit a scalability wall, they would have to learn design patterns to innovate once again and recreate their apps. They had to redesign the app with the new standards for modern widgets & components, testing it and feeling the app to see if they made the right choices. They had to show the app to their friends, family and random people to see how users would react to the application.

Of course, the more tech advances, frameworks emerge, and ease of programming is being installed, more and more people are at it. But AI sets the bar at such a low point that a literal child can operate an agent to create an app for them. The issue is that people that only know the feeling of a customer, that tries to build an app, often will fail to attempt to learn concepts needed to build the application. Just go on vibecoding forums and you’ll see a bunch of people arguing with the AI rather than trying to understand whats going on. This new wave of builder often ignores design, conception and testing, because they either let AI do it, or completely undermines these concepts and simply ignore them.

Ignoring design will eventually let the AI choose the same fonts, gradients and values they use for other people. It may follow Jakob’s law for ux, but wont create a unique experience.

Ignoring conception will let the AI use already existing solutions, that wont take into consideration the effort of thinking about the process.

Ignoring testing will prevent us from actually feeling the application and put ourselves as the customer.

I like the analogy of the car. A $400 000 supercar in the hands of an unexperienced driver might get him a faster lap than in a $20 000 econobox, but it wont get him to actually get better performance on a stage where people train to perform. But sometimes, if that person needed to get 15 minutes early to the hospital for an emergency, or if they just wanted to feel a small thrill, thats good enough for them to have the supercar to be faster. Skills are never defined by the toolset, but a good toolset will elevate someone with skills.

What can we do in this small world

Since AI is here to stay, there is a huge value into learning how to properly use AI, but not only that, but also how to master your own skills. If AI really multiplies the output of talented individuals, there is still a huge potential for agentic tools to get even better than they are. These tools still rely on competent individuals, as every missed key detail will result in a generic, suboptimal output. The skills needed to use AI effectively are learnt with experience, wether its the ability to prompt and use models properly, or your own abilities, that you train to use with AI leverage.

Conclusion

Experience cannot be replaced by AI. Your skillset is what defines you, makes you unique, and is what you will need to leverage AI properly. Even with the most advanced models, there will be a difference of output quality between an inexperienced person and a master. And remember, even if your AI model made the output, you are responsible for it.