Flagship · Pre-Launch / Internal Beta

Primestein

Thinks outside your box. Holds 9 possibilities at once. Never collapses to one answer until you need it to.

What It Is

Primestein is the divergent reasoning component of the Aurum Sphere substrate. Where Nomad talks like an LLM, Primestein thinks differently from one. It holds ideas in persistent uncollapsed superposition — 8+1 = 9 concurrent states, none collapsed to a single belief until explicitly needed. It is fed deliberately alien inputs: random arXiv papers, adversarial counter-questions, cross-domain bridge prompts from the swarm's 35 domains. The output isn't a reflection of the user's existing patterns. That's the point.

Primestein is why this system is different from another local LLM wrapper. LLMs reflect the curator's bias back at you. Primestein surfaces the thing you wouldn't have thought to ask. It runs autonomous multi-round synthesis with progressive trust gates — every action requires human approval, building a decision log that becomes training data for future autonomous operation. V0.1 is live in the current Sovereign Bloom substrate. Gateway mode: Brian approves every action.

Stage

Pre-Launch / Internal Beta. Primestein V0.1 is live inside the current Sovereign Bloom V5 build on the founder's hardware. Gateway mode active — every action human-approved. Not yet exposed to external users. Targeting CUDA + cuQuantum for production superposition compute.

Patents (Provisional — USPTO)

All patents are provisional filings through Aurum Nebula LLC. USPTO provisional → utility conversion deadline: March 2027.

Technology Stack

CUDAMulti-state superposition compute across 9 concurrent states
cuQuantumQuantum-adjacent hardware path for non-collapse architecture (Patent 056 OmniPU pulse)
DGX H100 / H200Production scale for full Sovereign Bloom deployment

Current: Rust (swarm_pulse binary, super_grid.rs), Node.js orchestration, Anthropic Claude API for scaffold reasoning.