The Ara Loop

Ara operates through a continuous improvement loop:

Find work → Ship it → Learn → Repeat.

Each pass through the loop delivers a real improvement and helps Ara perform better on the next run.

​

1. Finding the next improvement

Ara reads your product and codebase to determine what should be built next.

It looks for the highest‑value improvement and turns it into a runnable task. The work becomes a scheduled agent or background run, allowing progress to continue without waiting for manual prompts.

​

2. Shipping the change

Ara executes the task end‑to‑end. During a run it can:

• Edit files in your project
• Use the browser
• Use the terminal
• Open a pull request
• Review the resulting diff

The goal of each run is to ship a concrete improvement rather than only producing suggestions.

​

3. Learning the method

After completing work, Ara keeps the approach as a skill. Each run helps refine how Ara performs similar work in the future.

​

4. Preparing the next run

Ara stores what it learned and what was shipped. This memory improves its understanding of your codebase and informs the next improvement.

The result is a self‑driving loop where each iteration builds on the previous one.