Launching today
CANDOR.md
Open standard for declaring AI usage in software projects
8 followers
Open standard for declaring AI usage in software projects
8 followers
An open standard for declaring AI usage in software projects. Like a LICENSE or README, CANDOR.md makes AI involvement transparent and consistent. You can go with a global declaration, or a more granular apporach with parts of the software development process each having their own level of AI-involvement declaration. You can also go by component. The file works simply by using YAML front-matter enabling future tooling. The spec only defines the levels, the processes and the examples.
Prosaic
For a while now, across subreddits, social media and IMs, I've been noticing a trend of direct dismissal for AI-usage in coding projects even if the projects have merit and are made with the best of intentions. Along with that, I have been noticing a sort of dishonesty in the "I made this" when the person did not make the thing, does not truly possess the skills to verify what was made and has no way to declare this in a candid way.
In my understanding, there is a gap of candor here. And that is why I thought of CANDOR.md: An open standard for declaring AI usage in software projects.
CANDOR.md works without getting in your way, similar to a LICENSE file that you put in a repository. The standard does not have any major restrictions and only works as convention. It simply defines six (6) levels of AI-usage as well as different software engineering processes. You can declare a global level for each project or a more granular component-based or process-based declaration.
Think Conventional Commits but for declaring AI-usage.
The idea is to make CANDOR.md somewhat of a consistent source of truth in all repositories. The CANDOR.md specification is under the MIT License and is open-source, accepting discussion and PRs. The website and the repo is in the comments. The spec aims to to make AI-use transparent and, frankly, now that is here to stay, acceptable and deliberate.
And since we are using a markdown file with YAML-frontmatter, there is potential for tooling to identify the attributes and use the information systematically.