human cognition

2 items

RESEARCHarXiv CS.LG·4/14/2026

Human-like Working Memory Interference in Large Language Models

This study investigates working memory limitations in Large Language Models (LLMs), finding that they exhibit human-like interference signatures. Pretrained LLMs show performance degradation with increased memory load and bias by recency, even though transformers can be trained to perfectly solve such tasks.

LLMs AI limitations Working Memory human cognition

ARTICLEDEV.to AI·4/21/2026

Two Years Later: What My "Second Brain" Project Really Taught Me About Productivity

After two years and 1,847 hours, the author's ambitious "second brain" project, Papers, spectacularly failed to achieve perfect productivity and knowledge retention. This failure, however, revealed brutal truths about productivity, knowledge management, and human cognition, challenging the initial optimism about AI's ability to perfectly organize personal knowledge.

human cognition productivity second brain personal knowledge management