[KongchangAI]
Product

Sick Memory

A memory plugin for AI characters. Through a unified hub, relevance-ranked retrieval and a Dream offline consolidation mechanism, it enables the AI to remember the user while controlling token costs.

Timeline (last 90 days)

Nov 6

Sick Memory is a memory plugin designed to let AI remember the user and reduce the token cost of repeatedly carrying old content.

Unverified50%
Nov 6

Sick Memory uses a unified hub that first distinguishes identity, drawing and the boundaries of what can be said; facts are not overwritten, but understandings can have versions.

Unverified50%
Nov 6

Sick Memory's memory mechanism sorts by relevance, prioritizing memories that are relevant to the current topic, repeatedly mentioned, or emotionally strong.

Unverified50%
Nov 6

Sick Memory draws on the idea of how operating systems handle terminals, judging when a new message arrives whether the current task should yield.

Unverified50%
Nov 6

Sick Memory's Dream mechanism organizes memories offline by importance within budget, merging duplicate content and updating outdated understandings.

Unverified50%
Nov 6

Sick Memory's daily queries do not make extra model calls just to find a single memory, which is the key design for controlling cost.

Unverified50%
Nov 6

Sick Memory's persona is formed based on verifiable accumulated behavior (such as admitting not knowing, correcting mistakes, and protecting privacy), rather than a single sentence setting.

Unverified50%
Nov 6

In the usage records shown by the UP host, the cost screenshots include 5.59 yuan, about 2 yuan for daily use, and over 0.8 yuan for high-intensity use via a relay station, and the account is settled in RMB.

Unverified50%
Nov 6

The UP host explicitly states that the different cost screenshots were taken under different conditions and cannot be directly calculated as Sick Memory reducing costs by a certain percentage; there are currently no screenshots under the same conditions.

Unverified50%
Nov 6

Currently, Sick Memory's memory library and memory star map can be experienced through the Estabot and QQ paths.

Unverified50%

All Facts (10)

Unverified

Sick Memory is a memory plugin designed to let AI remember the user and reduce the token cost of repeatedly carrying old content.

50%
Unverified

Sick Memory uses a unified hub that first distinguishes identity, drawing and the boundaries of what can be said; facts are not overwritten, but understandings can have versions.

50%
Unverified

Sick Memory's memory mechanism sorts by relevance, prioritizing memories that are relevant to the current topic, repeatedly mentioned, or emotionally strong.

50%
Unverified

Sick Memory draws on the idea of how operating systems handle terminals, judging when a new message arrives whether the current task should yield.

50%
Unverified

Sick Memory's Dream mechanism organizes memories offline by importance within budget, merging duplicate content and updating outdated understandings.

50%
Unverified

Sick Memory's daily queries do not make extra model calls just to find a single memory, which is the key design for controlling cost.

50%
Unverified

Sick Memory's persona is formed based on verifiable accumulated behavior (such as admitting not knowing, correcting mistakes, and protecting privacy), rather than a single sentence setting.

50%
Unverified

In the usage records shown by the UP host, the cost screenshots include 5.59 yuan, about 2 yuan for daily use, and over 0.8 yuan for high-intensity use via a relay station, and the account is settled in RMB.

50%
Unverified

The UP host explicitly states that the different cost screenshots were taken under different conditions and cannot be directly calculated as Sick Memory reducing costs by a certain percentage; there are currently no screenshots under the same conditions.

50%
Unverified

Currently, Sick Memory's memory library and memory star map can be experienced through the Estabot and QQ paths.

50%

Source Articles