InsightInsight from our research and the field

All articles

Tech Note2026.09.21

Kubernetes Three-Node HA Setup and Failover Testing

Testing pod recovery in a three-node K3s cluster: local-path volume constraints, migration to shared storage, recovery results, and validation limits.

박예원

Tech NoteLLM Inside2026.09.17

Why is a long context window not memory? (Part 3)

A context window is working space for one request. Separating parameters, chat history, memory, and RAG shows why selection matters as much as length.

김해수

Tech NoteThe Ontology Diary2026.09.15

The ontology build and search improvement journey, Part 1 — Why we removed the LLM from the automatic build

How we separated the LLM from the automatic ontology build and found silent document loss, damaged Korean meaning, and false paths created by value nodes.

김진수

Tech NoteLLM Inside2026.09.10

What changes when an LLM has more parameters? (Part 2)

Parameters are trained numbers, not stored facts. Model size only makes sense alongside data, compute, post-training, and inference cost.

김해수

Tech Note2026.09.04

For voice AI, operating criteria matter more than recognition accuracy

How we connected live speech input to XGEN, displayed interim recognition while the user spoke, and handed off a request only after the utterance was complete.

Somin Park