# Taxonomy — OpenTentacle Labels v1.0

Full reference for all valid label combinations.

## Valid Level Codes

| Code | Name | Intent |
|---|---|---|
| `H` | Human | Creator used no AI tooling at any stage |
| `A` | Assisted | AI helped with non-creative, mechanical tasks |
| `C` | Collaborative | AI generated substantial drafts or assets; human shaped the final result |
| `G` | Generated | AI produced the core output; human chose or lightly polished it |
| `X` | Autonomous | Pipeline or agent ran end-to-end with no meaningful human review |

### Distinguishing A vs C

- **A (Assisted):** The work would look the same without AI. The human did the creative lifting.
- **C (Collaborative):** The work is materially shaped by AI output. Removing the AI contribution would require significant rework.

### Distinguishing G vs X

- **G (Generated):** A human reviewed, curated, or made deliberate selections from AI output.
- **X (Autonomous):** Output was published directly from a pipeline or agent with no substantive human review loop.

---

## Valid Type Codes

| Code | Covers |
|---|---|
| `txt` | Written language — articles, posts, captions, scripts, subtitles |
| `img` | Still images — photography (AI-enhanced), illustrations, graphics, UI assets |
| `vid` | Video — footage, animation, synthetic video |
| `aud` | Audio — voice synthesis, music generation, sound design |
| `code` | Source code, configuration, queries |
| `mix` | Content that blends multiple types where separating them is impractical |

---

## All Valid Combinations

Any `<level>-<type>` pair is valid. Common examples:

| Label | Meaning |
|---|---|
| `OTL:H-txt` | Article written entirely by a human |
| `OTL:A-txt` | Article written by human, AI-assisted grammar/spell checking |
| `OTL:C-txt` | Article co-written with an LLM, substantially edited by human |
| `OTL:G-txt` | AI-written article, human reviewed and selected |
| `OTL:X-txt` | AI-written article published without human review |
| `OTL:H-img` | Photograph taken and edited by human |
| `OTL:A-img` | Photograph with AI-enhanced background or upscaling |
| `OTL:C-img` | AI-generated image with significant human prompt iteration and editing |
| `OTL:G-img` | AI-generated image selected from a batch |
| `OTL:X-img` | AI-generated image published automatically |
| `OTL:C-vid` | Video with AI-generated segments, human-edited timeline |
| `OTL:G-aud` | AI-generated music track, human-curated |
| `OTL:C-code` | Code written with AI pair-programming (Copilot, Claude), human-reviewed |
| `OTL:X-code` | Code generated and deployed by an autonomous agent |

---

## Compound Labeling

Use the `+` operator when a single piece of content has multiple components with different AI involvement levels:

```
OTL:H-txt+G-img+A-aud
```

Use space-separated labels only when labeling **independent** pieces of content on the same page, not components of the same piece:

```
OTL:H-txt OTL:G-img
```

Do **not** use `mix` if you can separate the components — use `mix` only when the types are inseparably blended (e.g., a talking-head video where voice, video, and script are all generated together).

---

## Edge Cases

| Situation | Recommendation |
|---|---|
| AI generated a first draft that was rewritten entirely by a human | `H` — the final work is human |
| AI autocomplete was accepted occasionally while writing | `A` |
| >30% of sentences came from AI with light editing | `C` |
| Image was AI-generated but then heavily painted over by a human | `C-img` |
| Stock photo background replaced by AI inpainting | `A-img` |
| Podcast voice is 100% human but transcript was AI-cleaned | `H-aud`, `A-txt` |