Codifying human work
Half of all work in Western economies happens at a screen. Almost none of it has ever been written down. We are building the systems that can read it.
Conviction
When AI can build anything you can describe, the future belongs to the companies that truly know themselves
Work is the largest unrecorded dataset on earth
Every day, hundreds of millions of people move information between screens: reading, checking, copying, deciding. That knowledge runs the economy, and it disappears the moment the work is done.
Reading work is the hard problem worth solving
Work doesn’t arrive as data. It arrives as pixels, keystrokes and clicks, spread across a dozen tools, interrupted and picked up hours later. Solve that, and the knowledge that runs the economy stops disappearing.
- Screenshots
- Keystrokes
- Clicks
- Accessibility tree
- App events
- Capture
- Redaction
- Events
- Processes
- Model
- Company superintelligence
A company that understands itself can improve itself
Thousands of hours of work become one living model of the company: every team, process and exception, as it really happens. For the first time, a company knows how it actually works, not how it thinks it works.
AI should be onboarded, not installed
A new hire watches a colleague for a day and takes over one task. Onboarded well, they handle many within weeks, and the best ones don’t just do the work, they make it better. AI coworkers grow the same way, once someone can show them how the work is really done.
The point is people
We’re building toward a world where work flows naturally. The routine takes care of itself, and people have the space to imagine, decide and create.
Made with the time we didn’t have
We spend our evenings so everyone else gets theirs back. Sounds like you? See open roles