How To Build The Future: Aravind Srinivas

Perplexity CEO Aravind Srinivas argues Google’s ad-dependency prevents it from building the next generation of search, and that Perplexity is best positioned to win by obsessing over the user.

• Ilya Sutskever told Srinivas in ~2017 that only unsupervised learning + RL matters for AGI, redirecting his entire research focus.
• Srinivas and Ilya identified search and self-driving as the only two domains where product usage creates a flywheel that continuously improves the underlying AI.
• When Bing Chat leaked, one investor extended due diligence from 30 to 45 days and another asked for a call the next morning; Srinivas considered selling the company.
• Google Search generates ~$200B/year in revenue with the highest margins in Alphabet; Srinivas argues this stock-price dependency prevents Google from cannibalizing it with an AI-native answer product.
• Adding follow-up questions doubled engagement time on Perplexity and drove exponential daily query growth — the signal that convinced Srinivas not to pivot to enterprise.
• Perplexity’s core architectural bet: skip structured indexing and let smarter LLMs handle queries at inference time — a simpler approach that only worked once GPT-3.5-class instruction-following arrived.
• Srinivas’s long-term target: a single product that handles the full loop from question to fulfilled action (e.g., researching and buying a watch in one session), which he says is the real moat against OpenAI and Anthropic.

2025-02-21 · Watch on YouTube