Pencil versus prompt
I've always been interested in what ideas other people have to create open-source software. There can be many reasons why something is created, perhaps because you had a good idea, or some pain you want to fix, learning something new, or just by accident.
In essence I think someone creates when driven by an idea or challenge and wants to make it happen. I've experienced this a few times myself in my professional and personal life. It seems to me that nothing is more satisfying that turning an idea into something concrete, something that might be used daily.
Creation a solved problem?
With the arrival of powerful LLMs it seems the creation part of the challenge is something that has become more accessible and perhaps easier for anyone. I'm not touching the discussion whether the LLM produces good code or great quality.
If you know your way around an LLM then creating something that works is doable and a lot quicker than humanly possible before. I guess that the logical outcome of this is that at some point the user that creates does not have to be a developer anymore. It needs to be a person that can describe the need and get an LLM to create it. Again, we aren't talking about production ready code, just about the ability to create.
This always reminds me of a great example coming from Formula 1 racing, a sport where (super-) computing and automation plays a huge part in the development of new cars. Formula 1 teams build amazing F1 cars using a digital approach. These teams use 3D drawing software, wind tunnels, Computational Fluid Dynamics, incredible amounts of data, and will utilise anything digital they can get their hands on to find another few thousands of a second of performance.
One of the exceptions to this, and one of the most renowned designers of Formula 1 cars, is Adrian Newey 1. When people talk about him, they always seem to refer to his inhuman capability to see air flow (around a car). He is also known for his favorite design tools, a huge piece of paper and a pencil. He and his design teams have built multiple championship winning cars over time. He knows what he is talking about, yet he never uses computers for his initial design ideas.
This does not mean that Newey design teams do not use these tools eventually (they do), but the creativity of a new designs start on paper with a pencil with Newey. Which is amazing in such a complex technically driven sport.
The real challenge
Which brings me to what I think is the real challenge. If LLMs solve the creation part of the process, then why aren't we seeing incredible creations storming the world? I try to keep an eye out for things that are created in software. If, for example, I scroll a bit through hacker news, I see a whole lot of things being created. For some reason most of the creations aren't new or even interesting ideas. It seems that a lot of the creations may be useful, but from a distance they are often rehashes of earlier creations, or tooling ideas to aid the creation process with LLMs.
On the front of Hacker News, I count 10+ posts about someone building something to harness LLMs, more than 5 posts about someone porting an old game to make it work on some device or another, an actual product release, and a number of interesting posts about people hacking into hardware and building amazing things.
There is a lot of, here is a data set that we can visualise now because I can use an LLM to visualise it, type of posts too. The barrier to create something is lowered every day.
What is missing, for me at least, is people that are using their creativity to create something novel, something unexpected, something we had not thought about yet. It's an interesting observation. If we can build anything we want, then why don't we build something new?
The one thing you can't outsource to an LLM
It seems to me that the complexity of the challenge is quickly moving from creation to creativity. Weirdly enough, it seems to me that creativity is under a lot of pressure from LLM use. If everyone is asking an LLM what to build, it is likely that we will end up building the same things. You can brainstorm with an LLM (I know I have), but it is futile to assume you can get truly new or novel ideas from that. LLMs operate on statistical probability and known patterns, these are by definition not novel processes.
No matter how strong the software creating process will become with the help of LLMs, in my opinion the real challenge remains creativity. Thinking up new ideas or novel approaches do not really require an LLM or other fancy technology. As Mr. Newey has shown for decades, all it really requires is a creative brain, a pencil and a piece of paper. I may be naive, but I don't see that process being replaced by machines anytime soon.
Personally, I'm not really a creative person. It seems my brain isn't wired for it. My wife is the creative one, she is an artist, loathes computers and always curses when she must use one (never for creativity though!). The one thing I learned from her was that whenever I have a problem, need to find a solution, or I am just looking for inspiration, I grab a piece of paper and a pencil. It is the best tooling for coming up with ideas, for understanding what a problem is about, or to describe what a solution might look like.
Creation may have become easier, for creativity it is better to go back to the drawing board!
