Use the single official Agent — VibeCoding — to create a robot Skill that runs and can be demonstrated in Simulation. From idea to reproducible simulation demo.
Robot development is becoming useful — greeting, companionship, trial packages, simple perception loops — but most people still get stuck at two ends:
AuraOS turns “listen / see / think / call tools” into a sellable robot operating system. Simulation provides a virtual world usable for pre-purchase trials and development. The strongest developer lever on the current platform is VibeCoding — a single Agent that turns natural language into loadable Skill drafts, which you then validate in Simulation.
This hackathon proves: when VibeCoding and Simulation are connected, a developer can go from idea to a demonstrable robot scene without first building a managed multi-agent collaboration layer or swapping multi-organ hardware.
Your submission must clearly show the full loop:
VibeCoding must be a real part of the creation process. It cannot be used only as a thin wrapper, a final notification channel, or decorative screenshots.
requires_organs and degradation notes).
Please read carefully. These constraints define the current platform capabilities for this event.
| Supported | Not Supported in this Hackathon |
|---|---|
| 1× VibeCoding Agent | Managed multi-Agent collaboration / “Agent rooms” |
| Simulation demonstrations of movement, camera, and scenes | Built-in multi-Agent perception + planning + safety suite |
Skill files, requires_organs, tool loops |
Multiple official Aura Agents running in parallel |
| BP spherical robot or custom simulation models | Multi-morphology / multi-organ body swapping as a judging path |
Do not claim managed Agent rooms, a second official Aura Agent, or multi-morphology platform capabilities. Track 3 teams must supply their own reproducible orchestration algorithm, configuration, and demo.
Choose one primary track. All tracks require proof of VibeCoding usage + a reproducible Simulation demo.
Combine a persona card, scene specification, and one or more bound, loadable Skills in a single contribution package.
Contribute both a dataset and a model, with VLA or world-model approaches preferred and lightweight models used as edge or fallback support.
Contribute a reproducible multi-Agent orchestration algorithm, protocol, or scheduling strategy.
The default robot in this hackathon is the BP spherical companion robot placed in an indoor living-room Simulation environment. You may also use a custom low-poly simulation model of your own design. Multi-organ morphology switching is not supported in this edition.
dataset_card.md + training config (if applicable)Important: Missing any of the core items (video, repository, VibeCoding proof, architecture diagram, or safety checklist) will disqualify the project from the main awards.
Effective use of the VibeCoding → Skill → Simulation loop
Clear storytelling and visible Simulation behavior
Solves a real robot scenario or developer workflow problem
Creative use of the single-Agent craft + simulation
Each track is evaluated against its own required deliverables. Track 3 must demonstrate reproducible orchestration in Simulation.
Prize pool and specific awards will be announced closer to the event (including track awards, special awards such as Best VibeCoding Craft, Best Data Loop, Best Skill Packaging, and potential partner awards).
Non-cash incentives may include fast-track internship opportunities, early access to the Skill Store, developer certification, and more.
Exact dates and timezones will be announced on the official event page.
Open to students, independent developers, ISVs — anyone interested in robots, simulation, Skills, or AI tools.
Team size: 1–5 people. Solo is welcome.
Use the signup form at the top of the page to get early access.
Rules & Policies: