Exponential Episode 30: Selective Disclosure

Hank Korth has watched a lot of distributed systems cycles come and go.

A professor in the Department of Computer Science and Engineering at Lehigh University — with an appointment in the business college and as the lead at Lehigh Blockchain — he has been thinking about networked systems since the late 1980s, when his database research examined how independently administered systems could transact across a network.

On this episode of Exponential, he joined the show to talk about where his students and research team are concentrating their attention now: zero-knowledge acceleration, onchain privacy that can coexist with regulation, and the agentic era beginning to take shape beneath the headlines.

Lehigh Blockchain is both a research lab and a talent pipeline. Korth runs a systems-heavy course on consensus, cryptography, and smart contracts, and a campus-wide seminar on application and policy co-taught with former CoinDesk CEO Kevin Worth.

Alumni have gone on to co-found decentralized exchange Dolomite, help lead development on Aztec's Noir language, contribute to Zcash's Tachyon, and land at Oracle, Google's blockchain group, and Chainalysis. "We've tried to do not the, here's some little narrow academic thing, but rather to do work that people would care about," Korth says.

The research agenda reflects that priority. One thread is accelerating zero-knowledge proof generation on parallel architectures, where Korth's team has found that memory — not compute — is the dominant bottleneck, pushing the work toward multi-GPU algorithms.

Another is performance benchmarking, first across layer-one chains and more recently across layer twos, including the surprisingly slippery question of when an L2 transaction should actually be considered final. A third, and the one Korth spent the most time on, is the intersection of privacy and regulation.

His premise is that absolute transparency and absolute privacy are both dead ends.

Fully exposed transactions leak a user's entire financial history to anyone they pay; fully shielded ones invite the Tornado Cash problem. Korth's group is building a framework in which a user submits a private transaction accompanied by a zero-knowledge proof of regulatory compliance — with "compliance" intentionally left as a parameter.

The long-term ambition is to get regulators, or trusted institutions acting on their behalf, to publish rules in a logical form that can be compiled, via Noir, into proof-generating code. In Korth's view, that creates a competitive dynamic: jurisdictions that publish proof-friendly regulations become more attractive places to do business, and stablecoin ecosystems start to compete on privacy features rather than the price of a dollar.

The same architectural logic, he argues, extends to agentic AI. An AI agent transacting on a user's behalf cannot open a bank account, so it will settle in stablecoins — and it will need to prove to counterparties that it genuinely represents its principal, has the resources to act, and is disclosing only what it must.

Korth also sees a decentralized alternative to monolithic LLMs beginning to form through projects like BitTensor, where small models interoperate across an L1 with incentives designed to reward collaboration without re-running the Web2 pattern in which, as he puts it, "my Facebook post is not mine, it's Meta's." Medical AI — where providers need to prove that an FDA-approved model produced a diagnosis without exposing patient data or proprietary parameters — is another domain where the same proof-attached pattern shows up under a different name.

Closing with a prediction, Korth reached for a reference most of his students would miss. Borrowing the famous career advice from The Graduate, he offered a two-letter answer for where the next decade of meaningful work will come from: "ZK is the new plastics."