3 posts
Stateless LLM calls aren't enough for agents. Short/long-term memory, episodic-semantic-procedural memory, and practical architecture in light of KVKK's Agentic AI guideline.
An agent at 90% in testing drops to 70% in production. A field guide to the reliability gap, pass^k, LLM-judge biases, and a production evaluation framework.
In 2026 agent memory is a layer separate from the context window. Short/long-term memory, Letta, Mem0, Cognee and KVKK-compliant production design.