Andrej Karpathy on

ai-architecture

2 entries, 10 Aug 2026 to 10 Aug 2026

On the recordsourced and dated, oldest first

    1. reported

      Large language models cannot retain new information told to them during conversations.

      “Current large models lack continuous learning. You can’t tell them something and expect them to remember.”

      10 Aug 2026 · kucoin.com · source
    2. reported

      Karpathy proposes a small model of one to two billion parameters combined with external memory that can grow over time.

      “ (which he estimates needs only one or two billion parameters) paired with a structured external memory system capable of self-reinforcing growth. Following this idea, he released a model called ”

      10 Aug 2026 · kucoin.com · source

Andrej Karpathy ontheir other subjects

Everything Andrej Karpathy is on record saying · RSS