Where we’re looking.

Intelligence that works for one person, on their terms.

We are interested in AI that runs where people are — on their own machines, with their own data, under their own control. Capable, private and dependable enough to trust with real work.

• Open questionHow small can a genuinely useful model be?
• Open questionWhat does an assistant look like when it never needs the cloud?

Software for the machines people do their best work on.

The desktop is where serious work still happens. We want to make the tools on it calmer, faster and more coherent — fewer apps, less noise, more done.

• Open questionWhat would an operating layer for focus look like?
• Open questionHow should tools share context without sharing your data?

Experiences built for the device that never leaves your side.

Phones hold our most personal information and our most fragmented attention. We see room for products that respect both.

• Open questionHow do you build for a device that is always with you, without demanding attention?

Protection that is invisible until the moment it matters.

Security should be a property of good design, not a feature bolted on later. We want privacy to be the default — and to be something people can actually see.

• Open questionHow do you make privacy verifiable, not just promised?
• Open questionWhat does safe automation look like?

Foundations other builders can stand on.

Over time, the most valuable things we build may be the ones others build on top of. We intend to share what we learn as infrastructure, not just as products.

• Open questionWhich of our internal tools would be valuable to everyone else?