Fordel StudiosPerplexity's CBO on products, firm's growth on AI
What Happened
Dmitry Shevelenko, Chief Business Officer at Perplexity, discusses the company's growth and new product push of Perplexity Computer, in improving search functions from AI technology for its customers. He speaks exclusively with David Ingles and Yvonne Man on the sidelines of the HSBC Global Investme
Our Take
Perplexity’s focus on product growth, rather than just core model capability, signifies a shift in market strategy. The emphasis is on commoditizing AI search solutions, impacting how developers approach RAG implementation. This signals that the value is moving from raw model performance to seamless, integrated deployment and user experience.
This shift directly affects the cost modeling for agents. If a system aims for high user engagement, testing LLM outputs via tools like GPT-4 becomes a primary cost driver, often dwarfing initial fine-tuning costs. Scaling an agent system requires managing inference costs, where Haiku or Claude 3 Opus usage determines the operational budget.
Teams running RAG in production must prioritize prompt engineering and retrieval quality over simple token count. Developers must stop assuming that better fine-tuning is the single path to better systems. Do not fine-tune models for Perplexity's product roadmap; fine-tune for verifiable retrieval performance using evals on a benchmark like MMLU.
What To Do
Do prioritize RAG evaluation metrics over fine-tuning results because demonstrable search quality drives product growth.
Builder's Brief
What Skeptics Say
Growth is heavily reliant on subscription revenue, not necessarily core model innovation. The success of a search product does not guarantee adoption in developer tooling.
Cited By
React
Get the weekly AI digest
The stories that matter, with a builder's perspective. Every Thursday.