Unverified60% confidenceFactExact time
Traditional LLMs are "memoryless"—each API call is independent and the model itself does not remember the content of previous conversations.
2
Sources
60%
Confidence
Long-term
Relevance
8/29/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHy Memory uses a write-time LLM to convert each conversation into structured facts and reconcile them with old facts, whereas Aura does not use a write-time LLM and only retains the raw conversations75% similarUnverifiedLLMs are stateless, and every call requires resending the entire conversation history, causing the input tokens of a coding agent to expand approximately quadratically with the number of turns.73% similarUnverifiedLarge language models themselves are stateless, and each request is an independent computation; so-called 'remembering' usually relies on two mechanisms: one is concatenating historical conversation into the current prompt (the context window), and the other is retrieval-augmented generation (RAG) through external storage (such as vector databases or file systems).72% similarUnverifiedThe "forgetfulness" of large language models stems from their stateless inference architecture: each time the API is called, the model only processes the token sequence passed in the current request and retains no internal state across requests.72% similarUnverifiedLarge language models are essentially stateless functions, and a model's memory depends entirely on the content within the context window.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/817504API
curl https://kongchang.com/api/v1/knowledge/claims/817504MCP
get_claim(id=817504)