I was tired of repeating the product.
So I built a record the agents could share.

Why I started RAAV

I am Ville Ylläsjärvi. While building FeedSquad with coding agents, I kept losing the thread between sessions. The code was in Git, but the reasons behind it were spread across chats. I had no dependable view of what another agent was changing, which decisions were settled, or what had really been tested.

What kept breaking

Each chat looked productive on its own. Across the whole build, the gaps were harder to ignore:

• A new session had to reconstruct decisions I had already made
• Two worktrees could drift toward different versions of the same feature
• An agent's confident summary did not tell me which checks had passed
• Handing the project to another person meant another long explanation

What I am building

RAAV is a small project record shared by the founder and the coding agents. It does not write the code. It gives each run a known starting point and asks it to leave a useful record when it stops.

• The Product Handbook records the current product and open questions
• Tasks, branches, worktrees, and file claims show where work is happening
• Each completed run records files, checks, risks, and the next action
• Product changes wait in a review queue until the founder responds

Who I am building it for

The first user is a solo founder who can direct a coding agent but does not want to manage the product through terminal logs and Git internals. I also expect it to help small teams and developers inheriting agent-built code, but I am deliberately testing the founder use case first.

Where it stands today

RAAV is in controlled founder testing. I am checking the CLI, local project record, hosted console, and agent connector as one workflow before opening public checkout. The changelog records what has shipped, and the status page lists current limits.