With AI agents making it easy to build custom dev tools quickly, how should we think about releasing tools that are similar to existing ones? For example, tools like SoloTerm are great, but agents make it possible to recreate and customize similar tools in hours instead of months. Where do we draw the line between normal iteration and copying? I’ve built a few personal tools this way and haven’t wanted to release them publicly, which made me wonder how others feel about sharing things built with heavy AI assistance.

“Coping” is subjective term. If you understand the design and code and understand the pros and cons of the LLM generated output it should be fine. I/and WE (referring to lot of people in this industry) personally have copied a lot of things like from stack-overflow/google/blogs etc. One thing that i find missing in the current crop of LLM’s and the hype surrounding it is that it never provides an counter argument to the code/design that it produces. Before when you ask a question in stackoverflow you will get multitude of responses each providing the scenario when you should use one over the other. (There is NO one shoe fits all solution, but the LLM’s make you feel so.) So to answer your question, if you understand the code/design you should be fine, even though it might feel like cheating, as you can modify it and improve on it later.

One big think i draw upon is open source. I have seen folks just plainly copy open source code (btw this happened before the coding agents), but now people copy/recreate and then publish. Oss is not about only code, its about quality, maintenance and support. So we should try and maintain that with oss.