Show HN: AI memory with biological decay (52% recall)
4.2(10,000 条评价)
免费
2026年上线
Sobre
Most RAG setups fail because they treat memory like a static filing cabinet. When every transient bug fix or abandoned rule is stored forever, the context window eventually chokes on noise, spiking token costs and degrading the agent's reasoning.<p>This implementation experiments with a biological approach by using the Ebbinghaus forgetting curve to manage context as a living substrate. Memories are assigned a "strength" score where each recall reinforces the data and flattens its
优点
- +Implementação de uma abordagem biológica para gerenciar o contexto
- +Uso da curva de esquecimento de Ebbinghaus para gerenciar a memória
- +Reforço da memória com cada recall
缺点
- −Pode ser complexo de implementar
- −Requer ajustes para diferentes cenários de uso