Dots keeps your attention on what matters.
The desktop attention layer for you and your AI agents.
Why Dots
More agents can do more work. Your attention still has limits.
Work now spans applications, conversations, and agent providers. People spend attention piecing it together, while agents spend tokens rebuilding context. We’re building Dots to curate what each person and agent needs, and help determine when to help or stay quiet.
Three ways to help
Dots works alongside you and your agents, within the tools you already use.
01Ask DotsFind answers across your work.
Ask about earlier decisions or related work across applications and agent sessions. Dots brings that context together and shows the sources behind its answer.
“Did we already decide this in another conversation?”This illustrates a question Dots is being built to answer.
02Equip your agentGive your agent the context it needs.
Dots gives your coding agent relevant decisions, constraints, and prior work from other sessions. You can continue working without repeating the same background each time.
“What changed elsewhere that this agent needs to know?”This illustrates a question Dots is being built to answer.
03Dots speaks upDiscover connections that help you move forward.
Dots connects information across your apps, conversations, and agent sessions to find insights you might otherwise miss. It suggests useful next steps when they can help you make progress, without waiting for you to ask.
“Your support notes and onboarding feedback describe the same setup issue. Simplifying that step could address both.”This illustrates a suggestion Dots could make.
Attention Labs
We equip models and people with systems that learn where to direct their attention.
Dots is our first product. Its desktop form lets assistance stay quietly available across applications and agent sessions, so people can feel supported and in control. Our ambition extends to future research previews and products built around attention science.
Context helps a system understand what is relevant. Attention connects that understanding to behavior: what help is useful, how much is enough, and when it should reach a person or agent.
Our research
Attention is an alignment research problem.
People rarely spell out exactly how much help they want. Our research studies implicit intent and model behavior through real interactions, including what people act on, dismiss, correct, or return to.
We’re building systems that learn from real work and tacit outcomes. Our aim is to connect those signals to decision quality, rework, and progress, then use the evidence to improve what assistance surfaces and when. The goal is more useful work per unit of human attention and agent tokens.
Your data
You control what Dots remembers.
Dots keeps raw captures and personal history on your Mac. You can control capture, exclude applications, and ask Dots to forget information. We plan to add team sharing with your explicit permission.