Spec
An open-source full-size embodied AI robot dog project that aims to replicate Boston Dynamics Spot-level quadruped capabilities using consumer-grade hardware. It employs a three-chip architecture comprising a Raspberry Pi 5, Oak-D Pro W, and Jetson Orin NX, and supports fully offline operation.
Timeline (last 90 days)
Spec's currently publicly available information remains focused on the early development stage, and the stability of gait and stair-climbing, navigation robustness, and final actual cost have yet to be verified in subsequent updates.
In the Day 4 project update, the author demonstrated progress in gait and stair-climbing training as well as autonomous navigation, and revealed that most of the hardware had already arrived.
Spec adopts a divide-and-conquer architecture that decouples navigation, perception, and AI inference onto dedicated hardware to reduce single-point computational pressure and cost.
Spec is an open-source project aimed at building a full-size embodied AI robot dog comparable in size to Boston Dynamics' Spot but at a fraction of the cost.
The Spec project uses a Raspberry Pi 5 for navigation, serving as the main controller and motion planning core.
The Spec project uses an Oak-D Pro W depth-sensing AI camera for vision.
The Spec project uses an NVIDIA Jetson Orin NX edge AI module for AI computation and neural network inference.
Spec is capable of fully offline operation, with all AI inference completed locally on the onboard Jetson Orin NX without relying on a cloud connection.
Spec adopts a simulation-first sim-to-real strategy: motion policies are first trained in a simulated environment and then transferred to physical hardware.
Wayfinder works by building a map on which every decision that needs to be made is placed as a question; when fog is encountered, agents are dispatched to research or build prototypes to clear it, and the result is ultimately converted into a Spec document.
1 more timeline events
All Facts (9)
Spec adopts a divide-and-conquer architecture that decouples navigation, perception, and AI inference onto dedicated hardware to reduce single-point computational pressure and cost.
50%UnverifiedSpec is capable of fully offline operation, with all AI inference completed locally on the onboard Jetson Orin NX without relying on a cloud connection.
50%UnverifiedSpec adopts a simulation-first sim-to-real strategy: motion policies are first trained in a simulated environment and then transferred to physical hardware.
50%UnverifiedSpec is an open-source project aimed at building a full-size embodied AI robot dog comparable in size to Boston Dynamics' Spot but at a fraction of the cost.
50%UnverifiedThe Spec project uses a Raspberry Pi 5 for navigation, serving as the main controller and motion planning core.
50%UnverifiedThe Spec project uses an Oak-D Pro W depth-sensing AI camera for vision.
50%UnverifiedThe Spec project uses an NVIDIA Jetson Orin NX edge AI module for AI computation and neural network inference.
50%UnverifiedWayfinder works by building a map on which every decision that needs to be made is placed as a question; when fog is encountered, agents are dispatched to research or build prototypes to clear it, and the result is ultimately converted into a Spec document.
50%UnverifiedA Spec (specification instruction) is a mechanism that packages the mature experience of senior developers into reusable collaboration templates, used to lower the barrier to Vibe Coding.
50%