SaaS Won't Die Because of AI

Source: Dev.to

1. Software is still not dead

There are two important points, even with AI:

• Software has not disappeared (and we all agree with that easily).
• Code has not disappeared; AI simply helps write it. The code is still there.

If software still exists, and it still exists in the form of code, there will always be a need for people who maintain it—software engineers. It doesn’t matter if non‑programmers can create software too. DIY has not rendered carpenters obsolete, right? So software and its makers are both still in the game.

2. Solving yesterday’s problem today is not necessarily an accomplishment

When I was a kid, I created my first crappy game using Game Maker, a free tool written in Delphi by a university professor. Ten years before that, games were incredibly hard to write because they required much more coding experience and ran on limited hardware.

Just because I can easily do what took talent yesterday doesn’t mean that talent isn’t needed anymore. Those folks were creating even bigger games—ones that were not possible with Game Maker at the time. It’s like saying that since I can learn calculus in high school today, we don’t need people like Newton anymore.

3. Creating some software doesn’t mean one can create all software

Large language models (LLMs) can interpolate very well, so asking them to solve problems similar to ones already solved for ages is easy.

The true test is how one uses LLMs to solve new problems—extrapolating beyond the existing knowledge. That’s the kind of boundary all the good computer scientists and software engineers are pushing. Most others focus on doing less to accomplish the same thing rather than setting bigger goals.

4. LLMs are good actuators that still need tools

Saying that we don’t need SaaS anymore is like saying we don’t need operating systems because each person can write their own. The fallacy is thinking that LLMs “think” and can “do everything.” If that were the case, LLMs would improve on their own, rather than Anthropic or OpenAI employing large numbers of humans (many of them computer scientists and software engineers) to improve the model itself.

LLMs require tools that they can call, and these tools need to be efficient. The entire universe of software, algorithms, and data structures is what LLMs would activate, but those machineries must exist first. There will still be a need for better algorithms, data structures, designs, and even tacit knowledge digitized by humans.

For example, LLMs are pretty bad at design. Even if they are good, they still need good abstractions. An LLM that integrates deeply with Figma and uses all of its tools and design abstractions natively is a killer combination—but it’s great only because Figma is a powerful toolbox for an LLM.

Conclusion

In short, software’s here. Code is still here. Yes, the nature of a software engineer’s job has changed—but when hasn’t it? It’s similar to moving from assembly to higher‑level languages, leaving low‑level details to the machine.

As long as there are software engineers and software to create, there will always be specialists providing specialized software products or services, which will invariably be better than hobbyist work.

What do you think?