After invite-only early entry, the desktop app from the crew behind Brevis is now open to anybody on Mac and Home windows.
HONG KONG, November 11, 2026: Arca, a brand new desktop app from the crew behind Brevis, opened to the general public right now. Arca pays folks for the AI classes they’d in any other case throw away. It finds the session historical past their AI instruments already carry on their laptop, lets them select which classes to submit, strips delicate particulars on the gadget earlier than something is uploaded, and pays for the classes which might be authorised. Beginning right now, anybody can join with an e-mail tackle, with no invite and no waitlist.
A lot of right now’s information work runs by means of AI instruments. Folks steer a mannequin by means of a process, correcting it, throwing out what fails and maintaining what holds up. Each a type of classes leaves an in depth document of actual work: the back-and-forth, the useless ends, the repair that lastly labored. These data pile up by the hundreds on folks’s computer systems, and as soon as the window closes, they’re value nothing to the one who did the work. Arca offers that work a worth.
The way it works
Arca works with the AI instruments folks already use, together with Claude Code, Codex, Cursor and Claude Desktop. After putting in the app, customers see the classes Arca has discovered on their machine, previous ones included. They choose a session to add, and Arca anonymises it on their laptop earlier than it leaves, then returns a quote vary. Customers both verify, which sends the session to overview, or withdraw, by which case the add is deleted. Accepted classes are paid weekly.
Arca solely reads classes which might be already on the person’s laptop, and it lists every part it finds. Nothing uploads within the background, and nothing leaves the gadget till the person picks a session and confirms. Earlier than a session is uploaded, Arca strips names, file paths, account particulars and keys, and additional anonymising runs on Arca’s aspect.
“Each AI session is a document of actual pondering and actual work, and till now the one who did that work received nothing for it,” stated Michael, co-founder and CEO of Brevis. “Arca lets folks see what their classes are value and resolve for themselves what to share. We opened it in small batches first to get the expertise proper. Immediately it’s open to everybody.”
Constructed for actual work in any discipline
Arca is constructed for anybody who does actual work with AI, from code to contract evaluations, analysis and lengthy working conversations. Periods with Claude Opus and Fable, and with the newest GPT fashions in Codex, qualify right now, with extra to come back. Very quick classes don’t qualify, and each submission is reviewed earlier than it pays. Customers can join GitHub to strengthen their profile, which might elevate the worth of their classes, however it isn’t required.
Arca carries the precept behind Brevis’s verifiable computing work into a brand new area: issues you’ll be able to confirm are value greater than issues it’s important to belief.
Availability
Arca first opened on October 13 in invite-only early entry, admitting customers in small batches. With right now’s launch, the invite requirement is gone. Customers join with an e-mail tackle or a Google account and may begin instantly. Arca is free to make use of and out there now for Mac and Home windows at arca.market.
Via November 30, the Arca Launch Problem provides $10,000 in prizes on high of standard payouts for authorised classes.
About Arca
Arca pays folks for the AI classes they’d in any other case throw away. The desktop app finds the classes folks’s AI instruments already maintain, lets them select what to share, anonymises classes on the gadget earlier than add, and pays for authorised classes. Arca is constructed by the crew behind Brevis. Be taught extra at arca.market.
About Brevis
Brevis builds verifiable computing infrastructure. Its zero-knowledge proving expertise, together with the open-source Pico zkVM, lets purposes show the outcomes of complicated computations. Be taught extra at brevis.community.
Media contact
Nicolas Tang
[email protected]
