Arca launches publicly, compensates users for otherwise discarded AI sessions

After a period of invite-only access, the Brevis team's desktop app is now available to all users on Mac and Windows.

Arca, a desktop application from the Brevis team, launched publicly today. It compensates users for AI sessions that would otherwise be discarded. The app scans session history that AI tools store locally, allows users to select which to submit, redacts sensitive information on-device prior to upload, and pays for approved sessions. As of today, users can register with an email address without needing an invitation or joining a waitlist.

A significant portion of knowledge work today involves AI tools. Users guide a model through tasks, making corrections, discarding failures, and retaining successes. Each session generates a detailed log of genuine effort: the exchanges, unsuccessful attempts, and the solution that succeeded. These logs accumulate in large numbers on local machines, and once the session ends, they hold no value for the user. Arca compensates users for that work.

How Arca operates

Arca integrates with commonly used AI tools such as Claude Code, Codex, Cursor, and Claude Desktop. Once installed, the app displays sessions it has detected on the user's machine, including previous ones. Users select a session to submit; Arca anonymizes it locally prior to transmission and provides a price range. The user can then confirm to send the session for review or withdraw, resulting in deletion of the upload. Sessions that pass review are compensated on a weekly basis.

The app reads only sessions already stored on the user's computer and presents a list of all detected items. No data is uploaded automatically; nothing transmits off the device until the user selects a session and provides confirmation. Prior to upload, Arca removes names, file paths, account details, and keys, with additional anonymization performed on Arca's servers.

“Every AI session is a record of real thinking and real work, and until now the person who did that work got nothing for it,” said Michael, co-founder and CEO of Brevis. “Arca lets people see what their sessions are worth and decide for themselves what to share. We opened it in small batches first to get the experience right. Today it’s open to everyone.”

Designed for professional use

Arca targets users engaged in substantive AI work, spanning coding, contract analysis, research, and extended dialogues. Currently, sessions involving Claude Opus, Fable, and the latest GPT models within Codex are eligible, with additional tools planned. Brief sessions do not qualify, and each submission undergoes review prior to payment. Users may link their GitHub account to enhance their profile, potentially increasing session value, though this is optional.

Arca extends the concept underlying Brevis's verifiable computing technology to a different field: verifiable data holds greater value than data requiring trust.

Public availability

Arca initially launched on October 13 in an invite-only early access phase, with users admitted gradually. As of today, no invitation is required. Users can register using an email address or Google account and begin immediately. The app is free and downloadable for Mac and Windows at arca.market.

Until November 30, the Arca Launch Challenge offers an additional $10,000 in prizes beyond standard payments for approved sessions.