< >

how it works

QORPUS intercepts every critical AI decision before it reaches the user. The same prompt is sent simultaneously to four independent foundation models — Claude, GPT, Gemini, and a fourth rotated for diversity. Each response is scored against the others using a proprietary divergence algorithm.

When all four models converge on the same answer, confidence is high and the output passes through. When they diverge — even slightly — QORPUS flags the decision for human review, showing exactly where and why the models disagree. The divergence itself becomes the signal.

Ten specialised agents — the BLOOMING CONSCIENCE system — then verify claims, check for bias, challenge assumptions, and gate the output before it reaches the user. No autonomous decisions are made. Every output is auditable.

technical references

QORPUS is model-agnostic and vendor-neutral. It communicates with foundation models via standard API endpoints and requires no fine-tuning or model access beyond inference. The system runs on local infrastructure using Ollama for lightweight models and cloud APIs for frontier models.

The divergence scoring engine compares outputs structurally — not just semantically. It measures agreement across factual claims, reasoning chains, confidence levels, and risk assessments. Results are stored in a local audit log with full provenance tracking.

The architecture is designed for compliance. Every decision, every model response, every agent verdict is logged and retrievable. This is not a black box — it is a glass box with a paper trail.

pricing

QORPUS will launch with three tiers. A free tier for individual researchers and developers, limited to a set number of divergence checks per month. A professional tier for small teams and consultancies who need reliable AI oversight for client work. And an enterprise tier with dedicated support, custom agent configurations, and on-premise deployment options.

Pricing will be published when the platform opens for early access. We are committed to keeping the core divergence engine affordable — trust should not be a luxury feature.

waitlist

QORPUS is currently in closed development, validated against one hundred real-world safeguarding scenarios drawn from UK regulatory investigations. Early access will open soon.

If you want to be first in line, get in touch via the contact page at wi-genie.com. We are particularly interested in hearing from healthcare providers, local authorities, legal teams, and anyone working in sectors where AI decisions have real consequences.