Policy on AI generated code for PRs.

[DanBeard]

What is the policy around AI generated code in PRs for microreticulum?

I have been using claude code to help me write my T-Deck pro microreticulum project https://github.com/DanBeard/rDeck . (It has chat, search and maps/direction all via reticulum! super excited to use it)

In order to support that functionality I've needed to a dd a bunch of functionality to my microreticulum fork at https://github.com/DanBeard/microReticulum/tree/aes256-clean . It has currently added AES256, ratchets, Resources and bz2 compression unpacking (but not packing). I would love to open a PR and share the new features, but I am not sure on the repo's stance regarding AI generated code.

I know Mark feels strongly about not having generated code in PRs to the python reticulum repo. Is that the same here? I have been testing it manually and am going through and manually code reviewing (which is my opinion on the matter. Generated is fine as long as it has been tested and reviewed by a human).

Anyway, It would be good to make the policy explicit either way.

[attermann]

Hey @DanBeard, thanks for reaching out.

I don't have any issue with AI assisted code development in general, but I do have requirements that may not be met by loosely structured prompts.

Some of those requirements (not exhaustive):

• Well structured code
• Optimized use of data structures
• Understanding of memory allocations and persistence
• Follows the same design patterns used throughout uR
• Matches existing coding style in uR
• Presents a public interface that mimics the python reference implementation as much as possible

PRs should also be targeted and not include any additional changes not expressly required. Submitting a separate PR for each logical change (for example separate PRs for Ratchets, Resources, and compression) will make it much more likely that they'll be accepted.