This project has migrated away from github. If you are a human, you can find the new repo on our personal code site:
https://gorge.glyphic.com/mark/ <-- typo on purpose
Change the leading g to an f. Why? 'Cause I don't want bots following that link.
I didn’t open source my code to be mindlessly regurgitated by AI. I opened sourced it to welcome other programmers to help and to build upon my work and learn from it.
My pledge: If you are not a programmer, or are just starting out: Don't use AI and I will help you learn. I'll take the time to carefully answer all your questions and review your code and mentor you.
Comments, issues, and pull-requests generated by AI will be summarily rejected: If you didn't take the time to write it, why should I take the time to review it?
Because the business of "AI" (specifically LLMs) is unethical no matter how you look at it:
-
It has incorporated vast amount of human output without asking permission, or respecting copyright. (Remember Aaron Swartz? he faced 35 years in federal prison for far less!)
-
The super rich companies that are promoting it are far overselling what it can do; while at the same time willing to lose billions of dollars saturating the market, until it becomes ubiquitous and we all have to pay for it. They are already floating that it is "too big to fail" and expect government bailouts to subsidize them.
-
The economic and enviromental impact of AI datacenters has been willfully underrepresented to communities. For example, on going external water usage is ~25x what the data center uses directly. While they create many jobs during construction, they have only modest employment in operation.
Technically, LLMs are inheritly flawed as a tool for writing code. Code is not like human language, and details and precision matter. If an AI written description uses slighly odd language or gets a minor detail wrong, a human reading it will check that it is reasonable, and gloss over the error.
With code, a minor looking error can cause programs to misbehave, in often hard to find ways. When you run code, there's no "reasonable" check on what it does. You are relying on the programmer to have done that when they wrote the code. Using AI alienates the programmer from the code. If the programmer used AI because they weren't proficient enough, or didn't want to take the time to write the code in the first place, then they definitely didn't review it in depth.