Show HN: AGI-HIVE — A spatial OS where multiple AI models argue over geometry

Most AI tools generate a blind output and ask you to trust it. When generating physical geometry for manufacturing, trusting a hallucination is expensive.

We built AGI-HIVE, a multi-model spatial computation platform that assumes foundation models will hallucinate. Instead of relying on prompt engineering, we force the models into provable consensus.

How the architecture works

• Multi-Model Consensus. When you prompt the system, it routes to multiple models simultaneously. The UI branches physically to show exactly where their logic splits. You review the evidence and act as the consensus mechanism to merge to trunk.
• ScalarCAD™ (SDF Ray Marching). AI struggles with polygon meshes. We use Signed Distance Fields (SDFs) and multi-scale ray marching. It evaluates mathematically perfect CSG operations in the browser.
• VSE (Visual Substrate Engine). Particle-based spatial data visualization for real-time telemetry and state analysis.
• The Evidence Ledger. Every generation and model consensus is hashed using BLAKE3 to create a cryptographic evidence chain.

Traceability by default

We aren't selling a better chatbot. We are selling traceability and cryptographic proof for generative engineering. Built by MurkWorks LLC.

Stack

Next.js, Firebase, WebGL2, ScalarCAD™, BLAKE3.

Closing

Would love feedback on the ScalarCAD™ pipeline or the consensus gating mechanics. Soft launch Apr 25, public May 29.