A minimal, keyboard-driven annotation tool for rapid human feedback on text snippets.

Load a batch of text items, review each one, type your annotation, hit enter, move to the next. Annotations are saved as you go. That's it.

Gloss is designed for situations where you need a human in the loop — quickly tagging, labeling, or commenting on a set of text data. The interface stays out of your way so you can focus on throughput.

• Keyboard-first — navigate and annotate without touching the mouse. Enter submits and advances. Arrow keys or hotkeys to skip, go back, flag.
• Resilient persistence — annotations save on every submit. No "remember to export" anxiety. Data lives in validated JSONL files.
• Import/export — bring data in as JSONL (or CSV, plain text). Export annotated results the same way.
• Minimal UI — one item at a time, one text field for your response, a progress indicator. Nothing else.
• Generalized — not tied to any specific annotation task. Works for sentiment labeling, AI-text detection, content review, tagging, or anything where a human needs to react to text.

Reviewing text snippets to identify AI-generated writing patterns and provide human feedback on what sounds artificial and why.

• Runtime: Bun + TypeScript
• Frontend: React (served via Bun)
• Data layer: Validated JSONL files
• No database, no auth, no build tooling beyond Bun