We were once told that AI would change everything
The coordination tax
But for most teams, that promise still hides behind a chat box. AI made us faster, but it has not changed how we work together.
It just turned up the volume.
We write more, we ship more, we generate more output. And all that output has to go somewhere. Endless notifications. Long PRs. Bloated docs. We’ve quietly traded “doing fulfilling work” for “routing information between systems.”
This is the coordination tax, and it compounds with every new teammate, agent, and tool. The bigger the organization, the harder it gets. We need to quiet the noise and get back our time: Time to think, space to create, and room for connection, while AI handles the busywork.
The most valuable team knowledge rarely lives in a polished artifact, it lives in the trail of daily work; the specific path to a decision, the screenshare and verbal explanation, the comment that changed a plan, the nuance of a discussion thread.
We are building the secure infrastructure to capture, model, retrieve, and apply this high-fidelity context.
In practice this breaks down into a few unsolved problems:
Proactive assistance
We are replacing manual prompting with autonomous, proactive action. Highlight surfaces what matters to you and your team, keeps track of context, and lets you run agents and workflows across tools before you even have to ask.
Shared memory
We've moved beyond “search a pile of text.” We are building a shared, unified team memory where AI understands the evolving state of a project by organizing context from every person and tool into a collective context graph.
High-fidelity capture
We're evolving Highlight into an always-on presence that stays with your workflow (via screen capture, audio transcription, and indexing), turning human-computer interaction into semantic data with total user control over privacy. Users choose what's captured and when it's shared.
Scaling collective intelligence
Memory without agency is just text collecting dust. We provide an execution layer for multiplayer agents that's grounded in context, enabling developers to scale a team's collective intelligence across every tool and system.
Our plan
At the core, we need a foundation of universal context made accessible and useful to teams, so we’re starting there. Our plan is to:
Build a 10x context capture system and shared memory layer.
Use the shared memory layer to power a proactive command center for teams.
Launch a platform for developers to build multiplayer agents on top of our shared memory and command center.
Enable teams to deploy fine-tuned world models as agents that learn by observation.
We’ve already proven our core technology with hundreds of thousands of users, and are now in early access with organizations.
Come build with us
We are a team of systems thinkers and founders from places like Discord, Medal.tv, and Meta who are obsessed with product and user behavior. If you want to solve the hardest problems in context and autonomy, join us.
See open roles