Zero contamination. 1,443 training pairs. 184 entity-knowledge examples generated by autonomous AI agents. Kush V2 is the first Hotep model trained on AI-cleaned data and validated by a 7-gate quality pipeline.
The first Hotep model on Meta's Llama 3.1 8B. A new naming convention, a new base architecture, and the bridge between v14's failures and Kush V2's triumph.
How a 7-agent autonomous swarm transformed response quality from C+ to A-, and what the upcoming Kush V1 model means for sovereign intelligence.
V14 trained successfully, passed compilation, and served real users — then scored 25/100 on quality audit. How contaminated training data hid in plain sight.
We switched base models, copied hyperparameters, and lost 29 GB of compute in hours. V13 was our most expensive failure — and our most valuable lesson.
V11 was not a model — it was an infrastructure revolution. How building a hybrid scoring system and rigorous data pipeline laid the foundation for every model that followed.
From v6 to v10 — how rigorous persona scoring, the Kosmos discovery, and an advanced RAG pipeline produced our most culturally aligned model yet. Plus: the v11 training pipeline is ready.
Announcing hotep-llm-v6, our most aligned model yet. Discover how rigorous A/B testing and 812 training examples produced a sovereign AI with authentic Hotep cultural knowledge.