Protocol Roadmap

Establish credibility as a neutral standards body, build an early waiting list, and attract data contributors.

• Landing page: hero, manifesto, sample leaderboard, and how-it-works sections.
• Waitlist signup: data-ingestion beta registration.
• Documentation: explanatory pages covering spatial AI, world models, metrics, and protocol glossary.

Open the first ingestion pipeline to a small group of contributors, reviewers and AI developers.

DePIN Data Layer

• Crowdsourced ingestion pipeline: upload interface for LiDAR scans (.las, .laz, .ply), CAD files (.step, .stl, .obj), robot trajectory logs (.csv, .json), and geometry imagery.
• Privacy and sanitization engine: automated metadata stripping, face detection and removal, and IP sanitization before data enters the archive.
• Point incentives: reward contributors for verified, rare or complex edge cases. Points are a record of contribution during the beta; they have no monetary value, and any future use will be decided through governance.

Verification Protocol

• Automated validation: file integrity checks, duplicate detection, format verification and label consistency.
• Human-in-the-loop (HITL) review: independent reviewers will verify quality, correctness and labeling. Multiple reviewers per case, with agreement required before a case enters the benchmark.
• Reviewer incentives: points for accurate reviews, with reputation tracking.

Launch the live leaderboard and model sandbox API for independent testing.

Model Sandbox API

• Inference endpoint: AI developers submit model outputs (predicted meshes, point clouds, part labels, assembly trees) for scoring on verified held-out test splits.
• Evaluation metrics: 2D-to-3D construction accuracy, 2D/3D bounding box IoU, structural integrity, assembly tree accuracy, and physical logic checks. Detailed metrics are defined on the metrics page.
• Result publication: scores published on the public leaderboard with links to evaluation methods and code.

Live Spatial Leaderboard

Replace the current sample leaderboard with real model scores. Models ranked on verified real-world cases, not synthetic benchmarks chosen by their developers.

Transition to decentralized infrastructure and community governance.

Decentralized Storage

• Storage backend: evaluate Filecoin/IPFS or Arweave for large 3D asset files, CAD models and point clouds.
• Proof-of-Validation consensus: define criteria for verification consensus, data provenance, and archive integrity.

Non-Profit Foundation & Governance

• Foundation formation: establish a non-profit entity to steward the protocol, maintain infrastructure, and facilitate governance.
• Community governance: protocol changes (metrics, verification rules, archive policies) decided through open processes with input from contributors, reviewers and model developers.
• Point-based participation: contributors and reviewers will earn points during the beta. Any future use of points (such as governance voting) will be decided by the community, not by the foundation.

Long-term sustainability through optional enterprise services while keeping the core protocol free and open.

• Enterprise dataset licensing: premium access to high-fidelity, human-verified training data for AI labs fixing model hallucinations or sim-to-real gaps.
• Sky3 Compliance and Safety Certification: paid audits providing rigorous physical safety and reliability verification for robotics, autonomous vehicles, and industrial automation software.
• Custom benchmarks: private evaluation suites for companies testing proprietary models on verified real-world data.

These services fund the protocol without charging for core access. The public leaderboard, verification protocol and evaluation code remain free and open.

What's Not on the Roadmap Yet

The protocol is in its early stages. Several questions remain open and will be refined through the beta:

• Specific metrics for temporal prediction accuracy (world models forecasting motion over time).
• Task-specific scoring for manipulation, navigation and assembly planning.
• Robustness metrics measuring performance under sensor noise, occlusion and adversarial conditions.
• Material property recognition (rigid, flexible, brittle, transparent).
• The exact form of point-based incentives and whether they transition to governance weight.

These will be shaped by feedback from beta participants and the research community. All changes will be documented, versioned and decided through the governance process.