What Is ACM?

ACM is a system for running AI work in a way that you can repeat, compare, and review later.

Most people start with a simple problem. They have a document, a prompt, or a task, and they want to know which model or workflow gives the best result for a reasonable cost. Doing that by hand gets messy very quickly. You lose track of settings, forget which model you used, and end up comparing outputs that were not produced in the same way.

ACM exists to fix that. It gives you one place to set up a run, save the settings, execute the work, and keep the results. Instead of trying one thing at a time and guessing later, you can run a structured test and come back to it.

The goal is not just to generate text. The goal is to learn something useful from the run.

That might mean:

• finding the best model for a job
• seeing whether one workflow is more reliable than another
• checking whether the extra cost of a more expensive model is worth it
• scoring outputs instead of just eyeballing them
• saving a good setup so you or your team can reuse it later

ACM can be used in a very small way or in a much bigger way.

The small use is simple. Pick one preset, one document, and one model. Run it. Look at the output. That is enough to get value from the system.

The bigger use is a full pipeline. ACM can generate outputs, evaluate them, compare them, combine them, and save the whole job history. That is where it becomes more than a one-off tool. It becomes a way to run the same kind of work again and again without losing track of what changed.

If you want the shortest path to a real result, read quickstart.md.