AI Usage

How artificial intelligence is used in this project.

I’m a product designer amd I'm learning Rust, GTK4, and Linux desktop development for the first time through this project.

I studied Computer Science Engineering, began my career as a software developer in 2013, and transitioned into product design in 2016. Since then, coding has mostly been a side pursuit rather than my primary craft.

Since 2009, I’ve been releasing software projects publicly (imperfectly, curiously, and mostly for fun). Hangar continues that tradition.

I use AI throughout this project as an assistive development tool. This page exists to be transparent about how and where it’s involved. AI helps me think, refactor, debug, and explore unfamiliar territory but the direction, decisions, and responsibility for the product are entirely mine. Software is entering a new era of tooling. Pretending AI isn’t part of the process would be misleading. Overstating its role would be equally inaccurate.

How AI is used

Hangar uses Claude by Anthropic, primarily through Claude Code, as a development assistant. In practice, this means:

Some code is written entirely by Seth
Some code is written by Claude
Some code is written by Seth and refined by Claude
Some code is written by Claude and refined by Seth

This is a collaborative, iterative process. AI doesn't autonomously ship code to the project. It's a tool in the workflow, not a replacement. AI helps me explore unfamiliar Rust or GTK4 patterns, refactor messy code, debug issues, brainstorm architectural approaches, and translate ideas into working features faster.

Human oversight

Every piece of code that makes it into Hangar, regardless of whether AI assisted in writing it, goes through the same process:

Human review: all code is read, understood, and tested by a human before it ships
Manual testing: features are tested in real usage before being merged
Dependency scanning: the repository uses Dependabot for automated security and dependency updates
Open source: the full source code is public and auditable by anyone

What AI does not do

AI does not make design or product decisions
AI does not have access to user data (Hangar collects none)
AI is not running inside the Hangar application itself
AI does not commit, push, deploy, or release code on its own
AI does not make unsupervised changes to the codebase
There is no automated pipeline where AI writes and ships code without a human in the loop

Principles

Transparency Be honest about what tools are used and how

Accountability A human is responsible for every line that ships

Privacy AI is never in the data path between users and Bluesky

Auditability All source code is open for anyone to inspect

Questions?

If you have questions or concerns about AI usage in this project, feel free to reach out or open an issue on GitHub.