Cross-platform tracking
Watch mentions, snippets, and citations across major LLMs—mapped to queries and answer spans.
Desmond Li
NEOX R&D core · Ph.D.-led algorithms
Mattock models: capability scope and rollout cadence owned by Dr. Li’s team.
Dr. Li leads NEOX chip, Mattock models, and media-agent R&D alongside NEOXGEO GEO delivery.
Browse NEOX TechMattock models · GEO Core Model
Mattock models: brand visibility built for GEO
Mattock models are engineered for GEO; they embed architectures like the Mattock Neural Network and scan citations and share of voice across ChatGPT, Claude, Gemini, Perplexity—turning “what AI says about you” into work you can ship.
- AIO optimization
- LLM Visibility Index
- Attribution Audit
- Semantic Drift
Model signal preview
Use reverse prompt engineering to infer citation preferences and break visibility into an actionable AIO checklist.
platform.probemulti_llm_sync ✓
geo.visibility_index0.72 → target 0.91
audit.attribution_gap3 critical
Mattock Neural Network
Mattock: a scalable neural substrate
Mattock Neural Network is a core architecture of the Mattock stack—learnable activations and a forward structure that balance efficiency with scaling and distillation. Summary below; see PDFs at the end for detail.
Design focus
- · Scalable / growing-network mindset—approximate the target with a bounded parameter budget.
- · Multi-layer forward paths with learnable activations, aligned with the learnable-activation literature.
- · Validate on OCR / vision first, then extend to generative settings; distillation-ready.
Implementation & training (essentials)
- · Layer recursion with Hadamard products and matrix-valued activations for a compact forward map.
- · Each block: linear map + structured activation + diagonal composition (see PDF).
- · Classification heads may use Softmax; temperature controls sharpness.
- · Train iteratively with a loss (optional regularizers).
GEO core capabilities
From search… to being picked by AI
Public signals, on-site semantics, and multi-model answers on one scorecard—with LLM Visibility Index and next actions iterated together.
Gap diagnosis
Surface content discontinuities and semantic holes—the knowledge nodes to fill.
Competitor modeling
Compare how rivals rank and narrate inside AI answers.
Simulated console · GEO diagnostic
Visibility and citations on one screen
PRO dashboard mockup; live delivery is tailored by vertical and competitive set.
Five-axis signal map
Citation, trust, relevance, sentiment, authority
Multi-platform signals roll into a radar profile—fix what hurts LLM Visibility Index the most.
Interceptive analysis flow
Search to answer: intercept at the moments that matter
From search, retrieval, and summarization to the final reply—with Mattock intercept analysis and feedback loops marked along the path.
How it works
Reverse Prompt Engineering
Use probes and controlled experiments to infer evidence preferences and boundaries, rebuild “the prompt in the model’s eyes,” then turn that into AIO tasks—content, structure, proof, semantic consistency.
- → Infer ranking signals and citation thresholds
- → Align entities and knowledge-graph placement
- → Correct Semantic Drift in long-horizon narratives
- → Run Attribution Audit on how sources surface
Book a brand GEO diagnostic
Multi-model Mattock scans and AIO recommendations to grow visibility in generative search.
Book diagnostic