Skip to content

Power the future with your Mac.

The world is racing to build new data centers for AI. We’re bringing together the Macs people already own to help meet that demand.

Lithi connects those Macs with suitable AI work for businesses, from translating documents to reviewing code. Each Mac runs its part locally.

You choose when your Mac takes part. Once customers pay, 50% of eligible task revenue goes into a pool shared by the Macs that helped complete accepted work.

Our aim is to share more of AI’s value with the people and local communities helping power it.

“Before we build more, let’s make better use of what we already have.”

— Annette Abreu · Co-founder of Lithi
Months of development
Lines of code written
Active Macs reported in 2018

The people behind Lithi.

Annette Abreu and Ollie Hawkins met while traveling more than six years ago. Today they build Lithi together.

Annette Abreu smiling and holding a fan beneath a brick arch.

Annette Abreu

Co-founder · Chief of Integrity

Annette builds Lithi’s partnerships and community. She helps people understand how to take part.

Ollie Hawkins smiling from a palm tree on a tropical beach beneath a bright blue sky.

Ollie Hawkins

Co-founder · Mad Scientist

Ollie builds and tests Lithi’s technology. He turns experiments into software the network can use.

It started with a battery.

How a battery idea, a lot of AI experiments and one big cloud bill led to a network of Macs.

  1. The first idea

    A battery project gave Lithi its name.

    Our first idea was a power-tool battery designed to become useless if stolen. We named its software assistant Lithi, a play on lithium. The battery project didn’t last, but the name did.

    A yellow-and-black Lithi power-tool battery concept with an owner plate and QR code.
  2. Learning to run AI agents in parallel.

    Back in the glory days of unlimited token usage, Ollie ran hundreds of Claude Code agents in parallel. They wrote 1.2 million lines of code in one hour. Not all of it worked, but we kept experimenting.

    Ollie Hawkins beside two monitors and a laptop filled with terminal sessions during early parallel-agent experiments.
  3. Growing the network

    We built tools to introduce Lithi.

    Connectors are people who introduce Lithi to friends, neighbors and local businesses.What better place to start than over coffee at your local cafe? At a Community Hour, people bring their Macs, see how Lithi works and learn how to join. Connectors can earn from qualifying referrals when the people they introduce take part.

  4. A little help with the invites. And the replies.

    The Lithi app handles the busywork. A Connector picks a venue, and Lithi prepares the Eventbrite listing and posters. It also finds local Facebook groups and drafts posts and images to promote their Community Hour. The Connector reviews everything before it’s shared.People who see the posts and posters can text their questions about Lithi to the Connector’s local Lithi number. A personal SMS assistant answers and introduces itself as the Connector’s assistant. Less time figuring out what to say. More time getting people together.

  5. The cost problem

    Then we got the cloud bill.

    Our tests estimated more than $250 a month in cloud AI costs for one Connector using the tools at full capacity. At thousands of Connectors, the cash flow wasn’t going to work for us. Back to the drawing board!We started running local AI models on the Macs we already owned. We moved suitable AI jobs onto them and built software to make them work together.

  6. The result

    The Lithi Agent Network was born.

    The Macs running our internal tools became the foundation of the Lithi Agent Network. We kept improving the software, security and encryption. That work became Lithi Compute, for businesses with AI work, and Lithi Connect, for Mac owners and Connectors.

How Lithi works.

Several Macs can work on one job. Lithi divides supported work into tasks for approved Macs. It checks the results and returns them together.

Lithi divides supported work among approved Macs, checks the results and returns them together. Each Mac reads its assigned input. Dividing work limits how much one provider receives; it does not anonymize the data or provide confidential computing. The customer approves data transfer and a signed quote setting the price and maximum spend before paid work begins.

Work in

Lithi divides the work

Participating Macs

  1. Mac 01
  2. Mac 02
  3. Mac 03
  4. Mac 04
  5. Mac 05
  6. Mac 06
Checked results

The people and purpose behind Lithi

Still have a question?

Know someone with a Mac?

We’d love your help spreading the word. If someone comes to mind, please share Lithi with them.

Ollie & Annette