How To Build The Future: Aravind Srinivas

Key ideas

Srinivas traces Perplexity from enterprise search demos to a general answer engine, with "the user is never wrong" as a principle and queries per day opening every all-hands.

From a machine transcript of the audio. Quotes link to the moment they are said.

1. 1

   Pick problems where better AI improves the product

   Srinivas says he and Ilya Sutskever concluded that search and self-driving cars were probably the only problems where you can work on AI and product at once, because product use feeds data back into the AI. He adds that a product should keep improving as AI improves, so the company can keep going until AI is solved.

2. 2

   Demo first, then generalize what works

   Perplexity began with enterprise search over tables, but Srinivas says companies such as Crunchbase would not share data, so the team built a Twitter search demo in about a month. Trying to repeat it across domains led them to a more general design that leaves data unstructured and relies on the model at query time, betting that models would keep improving.

3. 3

   Let usage decide whether to pivot

   Srinivas says he was still unsure about the answer product even after people began sharing summaries of themselves. Adding follow-up questions doubled time on the site and raised daily questions, and he concluded it was not worth killing the product to pivot back to enterprise.

4. 4

   Treat the user as never wrong

   Srinivas credits Larry Page with this principle and says he keeps repeating it inside the company. In his example, when a query is ambiguous the AI should ask for clarification rather than say it does not know, and he contrasts that with products that teach users to become better prompt engineers.

5. 5

   Open every all-hands with queries per day

   Srinivas opens every all-hands with queries per day, reviews weekly and monthly growth rather than watching a screen all day, and shares the numbers across the company. He flags bugs directly to the engineer working on them and reads complaints on X because he finds people more honest there, though he admits the company is already slower than it was.

In Aravind’s words

What’s available here

• Original source

  Linked. Opens on YouTube.

• Player on this page

  Available. YouTube’s own player, loaded only when you press play.

• Full text or transcript

  Not reviewed. No license to show a text version here.

• Our summary

  Draft. Written by us with AI from a machine transcript of the video’s audio (Whisper) and checked against it. Machine transcripts can mishear; each quote links to the moment so you can hear it. Not yet rights-reviewed.

• Ask questions with AI

  Not reviewed. Off until there is licensed text to work from.