Festivus: The Open-Source Catalog for Physical AI
Normally on RoutineHub we talk about automation, shortcuts, AI workflows, productivity, and developer tools.
So yes, technically this project is a little off-topic.
But honestly, I think a lot of students, makers, and developers in the RoutineHub community will find this incredibly interesting.
Because Festivus is not just another AI project.
It’s an open-source project trying to build open infrastructure for the next generation of robotics and Physical AI.
And that deserves attention.
Open Source Infrastructure for Physical AI
Created by Haptic Labs, Festivus is an open catalog for robot policies, datasets, benchmarks, simulation environments, and deployment knowledge focused on the Physical AI ecosystem.
What makes it interesting is that it does not try to lock everything into a closed platform.
One of the best parts of the project is its philosophy:
“Papers stay on arXiv. Code stays on GitHub. Models stay on HuggingFace.”
Festivus simply connects everything together.
It does not try to replace the open-source ecosystem.
It tries to organize it.
The Real Problem With Modern Robotics
Right now, robotics is still surprisingly fragmented.
Policies are scattered across GitHub, HuggingFace, arXiv papers, incomplete documentation, abandoned repositories, and lab websites that disappear months later.
Datasets ship in different formats. Benchmarks are often impossible to compare fairly. And critical deployment knowledge is rarely shared publicly.
The Festivus website describes the problem perfectly:
“Nobody knows which policy ran on which robot, on which task, with which result.”
And honestly, that might be one of the most accurate descriptions of the current state of Physical AI.
More Than Just a Catalog
Festivus organizes the ecosystem into seven dimensions.
Robots
Hardware fingerprints, manipulators, humanoids, quadrupeds, and mobile bases.
Policies
Open-source robot policies including ACT, diffusion models, RL systems, and vision-language-action models.
Datasets
Robotics datasets in LeRobot and HuggingFace formats with provenance attached.
Environments
Simulation and real-world environments including MuJoCo, Isaac, and physical test rigs.
Tasks
Pick-and-place, locomotion, navigation, folding, and the shared vocabulary of Physical AI.
Papers
arXiv-linked robotics papers connected directly to policies and datasets.
Deploy Notes
Real deployment knowledge documenting what worked, what failed, and what the next team should try first.
The Most Interesting Feature: Deploy Notes
Probably the smartest idea in the project is Deploy Notes.
Because papers usually show:
- benchmarks
- metrics
- final results
But they almost never show:
- what broke
- what hardware failed
- what configuration actually worked
- what hacks were needed to make something run properly
And in real robotics, that information is often more valuable than the final benchmark itself.
Festivus wants to preserve the operational knowledge that normally disappears inside Discord chats, private notebooks, and internal lab discussions.
Why Share This on RoutineHub?
Because even if it is not directly related to Apple Shortcuts, it absolutely connects with something that has always defined this community:
Building Open Things on the Internet
RoutineHub grew because of:
- independent developers
- students
- makers
- people learning in public
- creators sharing open-source tools
- community collaboration
Festivus carries that exact same energy — just applied to robotics and Physical AI.
The phrase:
“Open Source Physical AI for the Rest of Us”
fits surprisingly well with the kind of internet culture many people here already understand.
An Invitation to Students and Developers
Even if you are not working in robotics, this project is still worth exploring if you are interested in:
- open systems
- AI engineering
- infrastructure
- datasets
- knowledge graphs
- developer tooling
- projects built in public
- open-source communities
Physical AI is still extremely early.
Which means this is exactly the stage where students, contributors, and independent developers can get involved while the ecosystem is still being built.
Why This Matters
A lot of today’s biggest technology ecosystems started with small open-source communities sharing tools and knowledge online.
Linux. GitHub. Stack Overflow. HuggingFace.
Physical AI will probably need the same kind of open collaboration to grow.
Shared infrastructure. Collective knowledge. Public deployment notes. Open compatibility between robots and policies.
And Festivus seems to be trying to become one of those layers.
Definitely Worth Watching
Maybe today this still feels far away from most of the RoutineHub community.
But a few years ago, it also felt unusual to talk here about AI workflows, LLMs, automation agents, and generative AI.
Now they are part of everyday life for many developers and creators.
Physical AI will probably be another one of those waves.
And open-source projects like Festivus are interesting precisely because they are trying to build the open infrastructure around that future from the very beginning.