The latest episode of Proof of Research is live! 🎙️ IC3's @sjbreck & @danivilardell77 join @mrrongxin to discuss their recent paper, πCreds, and how TEEs + LLMs are expanding the possibilities for privacy-preserving credentials. Traditional zero-knowledge circuits are great for simple checks (like proving age > 18), but struggle with broader semantic reasoning. By using LLMs and Trusted Execution Environments, πCreds allows users to prove complex claims without exposing their underlying data. 🎧 Catch the full conversation:

AI credentials are becoming an important part of global commerce and user verification. But what if the outputs aren’t always accurate? It could mean the difference between getting a loan or being rejected. Stress-testing how easily AI systems can be manipulated is how you build reliable technology. On Thursday’s episode of Proof of Research, IC3’s @danivilardell77 & @sjbreck talk about their work, πCreds, and why adversarial data attacks are an important thing to think about for LLM applications moving forward.

Will financial innovation dethrone the dollar, or make it stronger? In a new @FT op-ed based on his Jackson Hole paper, IC3 faculty @EswarSPrasad breaks down how stablecoins and blockchain tech could entrench USD dominance rather than dismantle it: ft.com/content/1481e787-77dc…

Incredible chance to interview Dani and Sam on their work with @AriJuels and thank you @initc3org!

In May, @flashbots_x, @initc3org, @theBBFund, and The Convent launched the Shape Rotator Accelerator in New York City. Over 10 weeks, it brought together people working across AI, hardware, privacy, creative technology, and cryptography to build for the agentic age under one roof. Shape Rotator was never meant to be a traditional startup accelerator. Set inside The Convent, a former nunnery and Catholic school, the space transformed into a workshop, lab, studio, and sound temple. Research ideas became prototypes, unexpected collaborations formed, and cross-disciplinary conversations turned into live experiments, culminating in a two-day finale where teams presented their work. The program was an experiment in storytelling: how do you make bold, technical, early-stage projects appeal to the rest of the world? It also demonstrated the deep curiosity surrounding what is being built at the edges of technology, where people can witness the process, not just the finished product. The accelerator teams shared the conviction that the most powerful technology emerges at the intersections between disciplines, software and hardware, AI and the physical world, engineering and culture. Thank you to everyone who built with us, experimented incessantly, and shape-rotated with us over these past 10 weeks.

Award for "Stealing ML Models via Prediction APIs" (USENIX Sec '16), led by @florian_tramer and @0xFanZhang, with Mike Reiter, @TomRistenpart, and me. We knew models were worth stealing. We had no inkling of the big impact the lessons on API design would have on frontier AI.

Congrats to Professor @AriJuels, Cornell Tech professor and co-director of @initc3org, on receiving the 2026 USENIX Security Test of Time Award! Learn more about his work: tech.cornell.edu/news/ari-ju… @Cornell_Bowers #CornellAI #USENIX