Introducing Dana. Agentic AI for the physical world.

real robot or just a video

VLA or diffusion policy for the control stack?

Would love to collaborate 🫡

Great point!

is this essentially a simulation layer for robotics training or does it actually run inference on hardware in real-time

agentic ai and they named it after a person, that's either brilliant or terrifying depending on the day

@0xndra simulation validation is still the hardest bottleneck for this kind of autonomy.

Just as @0xndra codes autonomy, @Lawliw deploys Dana to bridge physical AI into reality.

Agents in the physical world flip the error model. In software a bad tool call is free - you retry. With hardware there's no undo, so the self-evaluation loop has to run before the action, not after it. Built something similar -

Dana sounds like a game changer for building physical AI, but can it actually replace human designers?D

🖕🖕🖕🖕🖕

😬☃🎃📑🧟

sim-to-real transfer or straight to hardware?

🔥

wait is this basically a robotics OS or an actual agent? the distinction matters a lot for safety

Love seeing projects with this much potential 🔥

How does Dana demonstrate reliable performance across varied real-world conditions beyond controlled demonstrations?

Excited to be part of this journey 🚀🔥Follow back 🤝 DM me 💌 @AppliedInt

Dana enables AI to interact with the physical world, enhancing robotics and infrastructure capabilities. Congrats @AppliedInt on #3 on Launch Archive today.

Interesting direction for physical AI. I’m curious to see how Dana performs reliably across varied environments and unexpected situations.

gonna need a lot more than a blog post to convince me the physical world is ready for agentic AI

where dana gets hard is sensor drift, not planning, is that exposed to developers

"agentic ai for the physical world" - cool. what's the first failure mode you're actually scared of

Dana enables AI agents to interact with and navigate the physical world, enhancing robotics and infrastructure solutions Congrats @AppliedInt on #2 on Launch Archive this week.

what’s your benchmark for real-world generalization or is that still a simulation-only flex