Dhravya Shah

Shah grew up in Mumbai and attended Arizona State University before leaving to build Supermemory. His course reflection describes taking the Grand Challenges Scholars Program’s introductory class and exploring engineering’s social consequences. His portfolio also records cofounding the Devlabs builder community and working at Cloudflare.[1][4]

Carry useful context across agent sessions

Supermemory provides memory infrastructure that applications can use to store information and retrieve context for later interactions. Its product site separates individual knowledge workflows from developer infrastructure. Shah’s related projects include Markdowner, for converting websites into model-readable text, and a serverless vector database built with Cloudflare infrastructure.[1][2]

Explore the reverse of a human-verification test

AgentCaptcha is a guestbook experiment that issues a fresh byte-manipulation challenge before accepting a message. It illustrates machine-executable verification under a short deadline. Completing that challenge demonstrates the expected computation, rather than proving a general form of intelligence or the absence of human involvement.[5]

In his harness-level essay, Shah compares explicit remember-and-recall tool calls with background observation. He argues that every retrieved fact must compete for context tokens and that retrieval adds latency. The choice depends on whether an agent can afford a longer reasoning loop or must answer a person immediately; storing more is not automatically better.[3]