[KongchangAI]
Conceptlong-horizon memory

Long-term Memory

The ability of a system to retain and recall user-related information across multiple sessions and over timescales of days or weeks. It typically requires external storage and retrieval mechanisms and is a prerequisite for proactive agents to anticipate needs in advance.

Core Facts

Timeline (last 90 days)

Oct 24

The barrier to enterprise adoption of system-level agents lies in engineering aspects such as security (e.g., sandbox interception of dangerous commands like rm -rf /), memory (vector storage and long-horizon memory persistence), observability, and multimodal perception.

Unverified50%
Oct 21

Long-horizon memory refers to a system's ability to retain and recall user-related information across multiple sessions over timescales of days or even weeks, usually requiring external storage and retrieval mechanisms.

Unverified50%
Oct 21

The technical difficulties of long-horizon memory lie in write filtering, time-based decay, and privacy boundaries, which directly determine whether proactive behavior is thoughtful or intrusive.

Unverified50%
Oct 16

Two default practices for long-term memory: extraction should be placed asynchronously outside the reply path (because the facts of this conversation are already in this session's history); take a snapshot of memory at the start of each conversation (the memory block is placed before the history, and updating it midway would break the cache).

Verified65%

All Facts (4)

Source Articles