AI will definitely have some impact on the industry. I refuse to believe that the industry will one day cease to exist. Predicting what will happen is not only extremely interesting to think about, but also important if you want to stay ahead of—or at least in line with—the industry, avoiding pitfalls and bad practices that could harm you in the long run.

Not only do I not claim to know what will happen, but I actually don't believe most of the things I'm predicting here, either. I've explained in my book, A Philosophically Technical View Into Generative AI, why I think vibe-coding (delegating problem-solving to AI specifically) is a very bad practice. So naturally, I can't be optimistic about its future. But the reality is that the market has been convinced that vibe-coding is the way to go, and there's no turning back.

There are very different mindsets when it comes to AI. Each has its own set of predictions that are wildly different from the others. In this article, I aim for the middle ground: the positive-gray opinion, or in other words, "the most realistic best-case scenario." These guesses aren't well-supported, because if they were, I wouldn't have written my book the way I did. But I tried my best not to be shortsighted and to predict against my vision, seeing past my own flaws and into the future. Take them with a box of salt.

  1. Literacy and linguistic precision would become two of the ruling factors. People who can translate intent into words more precisely would yield better results. This has always been true in software engineering, but until now, you could get away with being less precise with words by translating intent directly into software or visual definitions, skipping the word-based language part.

  2. Visual software development tools would emerge: deterministic tools that abstract away syntax (but would most probably differ for each language) and show semi-high-level stuff such as connections, dependencies, module structures, and even variables and chunks of each function. These wouldn't be used to change the code directly (like visual programming languages do), but rather to provide a much more reliable way to interact with AI, review its output, and maintain agency over source code. This would make knowledge of coding (not software engineering) respected once again, though not in the same way as before.

  3. Programming languages wouldn't die, nor would they be developed much faster. A slight increase in development speed would be natural, but it would be smaller than expected. Liability and the need for reliability wouldn't allow a 10x increase.

  4. Deep knowledge of code would become a niche for two groups:

    1. Highly important jobs that carry much more liability, from OS development all the way to rocket programming.
    2. The "code medic" people, a new kind of freelancer who would fix broken workflows in other people's projects, debug their code, get them out of loop hell, and so on.
  5. Tutorials wouldn't die. Instead of learning how to code a website, you'd learn how to visually develop a website and how to translate intent into words and technical terms.

  6. Technical knowledge would become both deeper and higher-level. Developers would need a deeper understanding of the technical concepts they already work with to use visual development tools effectively. In other words, experience would become more important, not less. At the same time, the increased speed of development would shift more of the demand toward higher-level engineering skills.

  7. Software quality wouldn't get worse, but the average experience would. The new workflow would enable less knowledgeable people to generate stuff that's "good enough" and flood the market. Good stuff would still be discoverable for those who care, but the average experience of an average person would decrease in quality, unless some kind of awakening happened. The increase in the speed of prototype production would also allow good developers to shine more. Video games, in particular, wouldn't decline in quality but would actually improve, because they involve so many interconnected systems and details that you can't just throw one together without care and expect it to be "good enough."

  8. The barrier to success would actually increase due to instant copycats. You might enter the industry much faster, but competing would become harder, because you'd have to put more time and effort into your game, enough to make it increasingly harder to one-shot it. At the same time, this would push you to make better games rather than games built around individual mechanics, making it easier to sell your work if you actually care about game development as a career.

  9. Reverse-engineering would remain unimportant because piracy is almost always a service issue. Reverse-engineering could instead be used by developers to test their random ideas without even needing to prototype that much.

  10. Context would matter less. AI would become better at seeing connections and managing its own context, while the rest would be managed on a per-task basis by the visual workflow. Agent proposes, human reviews, agent implements, human verifies (not the end output, but the code output through visual tools).