# DinoDS

> DinoDS turns raw web data, scraped data, internal documents, and workflow examples into validated LLM-ready datasets for fine-tuning, RAG behavior, structured outputs, tool-use behavior, and agentic workflows. Train behavior, not just responses.

DinoDS is a B2B portal for action-oriented LLM training data. We compile behavioral datasets with schema-locked training corpora across a 37-lane taxonomy, so teams can teach models to *act* — not just talk. Delivery is operated through a Discord command center.

## Pages

- [Home](/): Positioning, format exporter sandbox, and lead capture for the DinoDS dataset compiler.
- [Lanes](/lanes): Full registry of the 37-lane action taxonomy with technical contract details.
- [Dino Lab](/lab): Interactive AI simulation sandbox to experience action-trained model behavior.
- [Lab Prompts](/lab/prompts): Curated prompt library for the Dino Lab sandbox.
- [Benchmarks](/benchmarks): Empirical performance dashboard showing model deltas before vs. after action training.
- [Methodology](/methodology): How DinoDS compiles, validates, and ships behavioral datasets.
- [Dispatch](/dispatch): Educational hub and editorial posts on action-data and agentic training.
- [Discord](/discord): Entry point to the Discord command center used for operational delivery.

## Optional

- [Unsubscribe](/unsubscribe): Email preferences management.