Gauntlet founder and researcher applying simulation and market design to blockchain systems and AI infrastructure. His writing examines how incentives can change the behavior of a technical system.[5][2]
Start with Caching Cheaters on OpenRouter →
X · tarunchitra Career
Before Gauntlet, Chitra worked at D. E. Shaw Research and Vatic Labs. His Stanford symposium biography describes high-performance-computing and AI simulation research, including ASIC work at D. E. Shaw in 2011. His Cornell senior thesis examined mathematical physics through geometric and topological methods.[5][6]
Contributions
A cheap token quote can hide a costly execution pattern
Caching Cheaters on OpenRouter examines a marketplace where providers advertise token prices but also report cache usage. If repeated requests receive no cache discount, a lower advertised rate can produce a higher final bill. Chitra compares repeated-query behavior and costs for providers of one model, then discusses market-design changes that could make this behavior visible. The empirical claim concerns the tested providers and workload.[2]
Study competition for the same security collateral
His staking-and-lending paper models validators who can either stake tokens to secure a network or lend them through contracts. Changes in returns and risk preferences can shift tokens away from staking. The stochastic model identifies conditions for a transition between those equilibria, connecting a financial incentive with the network’s available security collateral.[4]
Tools, writing & talks
Writing, research & talks
Start here
A first-person investigation of cache reporting, advertised prices and total inference costs.
Start with the repeated-query example in the caching essay and compare the quote with the billed cost. Familiarity with input tokens and prompt caching helps; the argument concerns infrastructure incentives.
Author[2]Original sourceA stochastic model of how lending returns can compete with proof-of-stake security incentives.
Author[4]