OntosAI
-
Knowhere Notebook and MCP let agents in Cursor, Claude, and Codex search a shared cloud document Brain. Read more.
5–8 minutes -
Production RAG quality is mostly evaluation and data/state work—not model swaps. Read more.
8–13 minutes -
Knowhere is post-parse agent memory—tree hierarchy, multimodal assets, graph, and agentic retrieval—not a MinerU replacement. Read more.
8–11 minutes -
Document agents need persistent world state, not just retrieved observations. Read more.
7–11 minutes -
A history of OCR from mechanical templates through deep learning to multimodal large models. Read more.
24–35 minutes -
Reflecting on Microsoft SkillOpt: is editing skill.md real skill learning or prompt optimization? Read more.
4–6 minutes -
An explainer of Microsoft Research’s SkillOpt: treat skill.md as trainable text parameters. Read more.
5–8 minutes -
Applying Agent = Model + Harness to RAG failures rooted in ingestion and structure loss. Read more.
7–11 minutes -
Flat fixed-window chunking causes many so-called hallucinations; structured document memory helps. Read more.
5–7 minutes -
Knowhere open-source launch: tree-like parsing, table fidelity, OpenClaw plugin, SaaS and self-host. Read more.
3–4 minutes
Blog 