Task Management & Automation¶
The task is the core unit of unique creative production. The structure of the task is standardized; each task is unique creatively.
This is the resolution of the "you can't standardize creativity" objection: we don't. The structure around the work is identical for every task; what happens inside the art phase is never touched.
Task phases¶
Every task runs the same stations. A gate checks the work at every station, so artists don't have to — and defects never travel downstream.
| Phase | What's automated |
|---|---|
| Start | Building folders, copying default files. |
| Build | Assembly of upstream dependencies, default settings. |
| Art | The art — nothing to automate. |
| Render | Render settings, output directories, farm submissions. |
| Review | File conversions, review uploads, status changes. The work is judged by looking at it. |
| Refresh | Update scenes with newly published data. |
| Publish | Tech checks, quality checks, manifest data. The work becomes a part others can pull. |
Implementation: see Cookbook — Recipes, which defines the phase recipes per software/task, and the run_task orchestrator in Cookbook IDE & run_task.
Plugins¶
Native integration in every digital content creation package (DCC) artists require. List them, because each name is its own integration: Blender, Maya, 3ds Max, Unreal, Substance, Premiere, After Effects, Resolve, ComfyUI…
The economics are the point: in a traditional pipeline, a single DCC plugin is six months minimum — not for the code, for the buy-in. "Supporting editorial" alone is a seven-figure endeavor. We hire artists for their skill — if we require them to change tools, production is delayed; so the pipeline meets every artist in the tool they already master.
Implementation: see Plugins — Software Bridges and Porting a DCC to Alchemy.
AI services (task-specific)¶
Task-specific AI models plugged in per task — swappable the moment something better ships. Distinct from AI Integration, which is the platform layer: a model serving one task is a station tool; the ontology-aware agent layer is infrastructure.
Proof¶
- A 2,000-click, four-hour delivery process became one button.
- A full broadcast series was re-rendered weekly, run by one artist.