According to @karpathy, AGI is on roughly a 10-year horizon he describes as a decade of agents, citing major remaining work in integration, real-world sensors and actuators, societal alignment, and security, and noting his timeline is 5-10x more conservative than prevailing hype, source: @karpathy on X, Oct 18, 2025. He is long agentic interaction but skeptical of reinforcement learning due to poor signal-to-compute efficiency and noise, and he highlights alternative learning paradigms such as system prompt learning with early deployed examples like ChatGPT memory, source: @karpathy on X, Oct 18, 2025. He urges collaborative, verifiable LLM tooling over fully autonomous code-writing agents and warns that overshooting capability can accumulate slop and increase vulnerabilities and security breaches, source: @karpathy on X, Oct 18, 2025. He advocates building a cognitive core by reducing memorization to improve generalization and expects models to get larger before they can get smaller, source: @karpathy on X, Oct 18, 2025. He also contrasts LLMs as ghost-like entities prepackaged via next-token prediction with animals prewired by evolution, and suggests making models more animal-like over time, source: @karpathy on X, Oct 18, 2025. For crypto builders and traders, this points to prioritizing human-in-the-loop agent workflows, code verification, memory-enabled tooling, and security-first integrations over promises of fully autonomous AGI, especially where software defects and vulnerabilities carry on-chain risk, source: @karpathy on X, Oct 18, 2025.

According to Andrej Karpathy on Twitter, a significant new paradigm—system prompt learning—is emerging in large language model (LLM) training, distinct from pretraining and fine-tuning methods (source: @karpathy, May 11, 2025). While pretraining builds foundational knowledge and fine-tuning shapes habitual behavior by altering model parameters, system prompt learning enables dynamic behavioral adaptation without changing parameters. For crypto traders, this development could accelerate AI-driven trading bots' adaptability to new market conditions, enhancing execution strategies and potentially impacting short-term volatility as AI trading tools become more responsive (source: @karpathy, May 11, 2025).