How To Build The Future: Aravind Srinivas

TL;DR

Perplexity CEO Aravind Srinivas describes the company's evolution from an enterprise-focused database search tool to a consumer AI search engine that uses LLMs to provide direct answers with sources. The core insight driving Perplexity is that as models improve, a simpler "dumb approach" of feeding top link summaries into the prompt becomes viable and fast, creating a product advantage.

Key insights

• Ilya Sutskever told Srinivas that only unsupervised learning + reinforcement learning matters for AGI, dismissing other research as useless – this shaped his focus on generative models.
• Srinivas realized after launch that Google cannot easily put a direct-answer product on its homepage because it would undermine the ad click economy, creating an opening for a new entrant.
• The first viral moment came from an academic searching herself – the model gave a past-tense biography because someone with the same name had died, sparking public debate about trust in AI answers.