The Problem
The founding thesis of Perplexity was blunt: search returned links, and links were not answers. In a 2024 conversation with Lex Fridman, Srinivas described watching users take four or five clicks to synthesise an answer they should have received in the first response, and concluding that generative models made a different default possible.
The commercial case was harder. Google had dominated search for two decades. Its distribution, ad-serving stack, and infrastructure lead were considered impassable. Every previous "Google killer" — from Cuil in 2008 to Neeva in 2020 — had failed either on quality or on economics. Srinivas and his co-founders believed the LLM era changed the underlying substrate enough that the incumbent's advantages did not automatically carry forward.
The bet was that the *interface* was ripe for replacement even if the *index* was not. Perplexity would build a citation-forward answer engine that acknowledged its sources, showed its work, and reasoned across web results rather than ranking them. If the interface was compelling enough, distribution could be earned rather than bought.
The Journey
Srinivas grew up in Chennai, India, and completed his undergraduate degree at IIT Madras. He moved to UC Berkeley for a computer-science PhD under Pieter Abbeel, one of the most cited researchers in reinforcement learning. Coverage in the Wall Street Journal has traced his path through OpenAI and DeepMind — internships and short-term roles that placed him inside the frontier research organisations building large models.
Responses
Join the conversation
You need to log in to read or write responses.
No responses yet. Be the first to share your thoughts!