NEW
SID-1
Our first agentic search model: 1.9x better recall, 24x faster.
Training SID-1 to beat GPT-5 at search with 1k+ QPS RL rollouts
Intelligence needs context

AI is giving humanity a new computer: one that can learn, reason, and act. But intelligence and skills are inconsequential without context. Context imposes the upper bound. At SID, we're training models to provide this context. To connect intelligence to the world.

Context needs search

To provide the right context, you need to find it first: Search is an ancient problem, but we think we have an angle to solve it once and for all. SID-1, our first agentic search model, doubles embedding-only accuracy and outperforms frontier models on the most complex search tasks.

Join us

We're a lean, We're looking for the ones that never settle. Instinctively. Persistently. Pathologically. Driven by the audacity to think "better" is always possible. And the resolve to make it so. The future is being built today not by the chosen few, but by those who make it their responsibility. What will you do with your one wild and precious life? team in San Francisco and Zürich. Backed by Y Combinator, Canaan, Rebel, and General Catalyst, as well as researchers from Anthropic, Deepmind, OpenAI, MIT, Cognition, Cursor, Applied Compute, Prime Intellect, Standard Intelligence, and Jeff Dean.

We’re always looking for great people to join us.

You can follow us on 𝕏.