Back in the day… that’s something people say when thinking about events that happened years, maybe even decades, ago. Well, maybe that was back in the day.

AI has advanced so quickly that this very post will already feel old by the time I finish writing it. It has become so easy to transform ideas into code, and then into applications, that there is no longer any meaningful time filter. We no longer need to spend days ruminating on an idea until we finally find time to start coding. We can simply write a prompt in seconds and have a prototype.

We, computer engineers, have become the bottlenecks.

Our minds no longer have time to recover between ideas. The previous prompt is barely finished before the AI agent is already asking for more work. They always want more, more, and more.

Back in the day, we would ask colleagues for help understanding a complicated idea, fixing a subtle bug, or improving the performance of an algorithm that had hit a wall. That required scheduling a call, or walking to their desk and breaking their concentration, explaining the context, keeping the conversation productive, and crossing our fingers hoping they could help us get unstuck.

Those days are fading away.

Now you probably have an AI agent integrated directly into your IDE. It has complete knowledge of your codebase, can track every change you make, and can begin any conversation already fully aware of what you are doing. That’s brutal. It is an area where human beings simply cannot compete. People need to talk, through speech or writing, and that takes time. Machines do that too, only incomparably faster, at least when everything works properly.

Of course, this reality is still being built. AI is still expensive for most people. The models are improving rapidly, but they still have noticeable limitations, and every decent programmer knows they cannot fully trust AI, especially in long-term projects. Its usefulness varies enormously depending on the field, which itself depends heavily on the data used to train the models.

And speaking of data, the origins of many closed-source frontier models remain questionable. Some people refuse to use them, claiming the data was obtained in a way no more ethical than how Aaron Swartz obtained his. Everyone knows how that story ended.

But let’s put all of that aside for a moment and imagine the world ten years from now.

Assuming no world war blows everything apart, AI costs will eventually become a solved problem. There is no way AI can continue consuming this many resources forever. Investors demand returns, and physical resources are finite. There is only so much silicon available to manufacture chips, and the world cannot dedicate all of it to OpenAI indefinitely.

The contexts in which AI becomes useful will expand dramatically. We will have extremely capable models, open source or not, running locally on personal devices. The so-called “Phase 3 of LLMs,” mentioned by Andrej Karpathy in his talk, will become reality for everyone.

There will be little reason to navigate settings menus or drag a mouse around anymore. You will simply say what you want and watch it happen.

“Change this color.”

“Install my printer.”

“Convert that video without losing quality.”

There will be no need to dig through mysteries you never truly cared about in the first place.

We do not learn new languages because we love grammar rules; we learn them to communicate with people who speak those languages. If technology can abstract away the language layer while still preserving meaning, then languages themselves may slowly lose relevance.

In this new world, what will it mean to be a Computer Engineer?

You probably won’t be a coder anymore. Nobody will need humans to write code manually, and perhaps nobody will even want them to, since machines will simply be more reliable at it.

Maybe we will finally become more like actual engineers.

We do not see civil engineers carrying stones around. They have machines for that. But those machines have no soul, no ambition, no personal problems they want to solve. Engineers do.

Perhaps we will finally be free from spending so much time figuring out the best way to move computation around, and instead focus directly on the problems we truly want to solve with that computation.

Maybe we spent so many decades talking about code, and calling ourselves “programmers,” that we forgot who we really are.

We are thinkers, artists, dreamers…

And perhaps it is time to remember that.

Pedro Alves