pushing code in the age of LLMs

We are deep in the “code is cheap” phase. Does it make any sense to publish code? For large software projects I believe it still does (it is cheap, but not so much), but in the realm of little tools it might not be such a case anymore.

What can be little code? Skills, extensions and cli tooling that we never accomplished because it was too much work for too little benefit. Not anymore, since code is cheap. Do these merit being published anymore, if given a prompt anyone can generate them? Is it worth the trouble and the attention it requires? And then again, taking this distribution / expression from me would be to kill part of the joy that I find in programming.

Even when little code is cheap, it might still be useful to other people. It sounds too forgiving to say “ideas matter, not code”, mainly because it just sounds as an excuse to accept the status quo, but to me it does make a bit of sense. Short programs, demos and things that glue things together are often more relevant for the idea they represent and not for how they are written.

But then again, how should I feel about pushing code I have not authored or even read? Because if it works, and you read it, it stops being cheap. The moment you put time on reviewing or making changes to it, it’s not cheap anymore. On a little piece of code, this is to me the turning moment when the agent owns the code and I own the original idea. I really couldn’t care less about that code, even how it is distributed, as long as it works and does not make me sink time into it.

Not sure if this is a thing or not, but this is what I am starting to do. Whenever I want to publish something that I have not really authored, it will come with a plus sign after my email username. That’s my way of signaling that I really do not feel ownership in any of that code, and it’s because it is both unfair and embarrassing to claim it as my own.

[user]
email = eskerda+ai@gmail.com
name = eskerda on AI