I came across Aartiq through the trailer and really liked the idea behind it. The Plan → Explain → Ask → Execute approach feels like a thoughtful way to build AI agents that can actually take actions.
What stood out most was the focus on permissions. Instead of treating AI capability as unlimited authority, Aartiq puts a clear step between what the AI can do and what it's actually allowed to do.
The demo of the browser agent breaking a larger research task into individual steps also makes the concept easy to understand. The permission and risk controls shown in the UI are another nice touch.
It feels like an interesting direction for AI-native browsers, especially for people who care about transparency and control.