AI Agents Can Orchestrate Local Creative Tools
An AI agent can turn mature applications on one computer into an adaptive production pipeline by choosing the right tool for each task and iterating towards a defined standard.
Developed from a conversation between Pete Winn and Andy David
Andy built a pipeline on his MacBook to create reels for the podcast. Instead of trying to recreate specialist features inside a new app, he let an agent use the tools already installed on the machine. It could work with DaVinci, Blender Headless, ImageGen and FFmpeg, selecting only the programs needed for each stage of the job.
The agent operated through what was essentially a scripted pipeline, but it retained discretion over which tools to use and could keep iterating until the output met a set standard. Andy found that this approach worked on its first run and produced a much better result than his earlier attempt to build every clipping and production function into a standalone application.
Pete saw this kind of computer use as a bridge rather than the endpoint. An agent can first discover a workable process through existing visual software, proving that the task can be done. Once the useful workflow is clear, developers can crystallise it into headless software that gives the agent direct access without the token-heavy overhead of navigating a human interface.