출처 시그널
a16z.news의 주요 글을 편집 맥락과 원문 링크와 함께 정리합니다.
a16z’s investment thesis treats AI inference as a stable, recurring workload where tokens per watt matter more than flexibility. Its sharper claim is that production—not chip design alone—is the moat: memory, power, cooling, interconnects, supply chains, and reliable rack-scale delivery determine unit economics.
AI agents break a core endpoint-security assumption: actions can no longer be attributed to the person at the keyboard rather than software acting with that person’s authority. a16z’s Neo thesis reframes control around real-time identity, agent guardrails, and attribution at execution; it remains an investment claim.
“World model” now labels systems that predict very different parts of reality. The useful distinction is functional: state, observation, action, and transition each support different capabilities, so visual plausibility alone does not establish physical understanding.