A technology executive, Computer scientist, Professor. A decade at Microsoft Research, executive roles at Brave, ZKSync, Eclipse.
Worldwide
Joined April 2015
- Tweets1.7K
- Following3.8K
- Followers8.8K
- Likes1.5K
Are we entering a golden age of software security, or just a faster loop of vulnerability generation?
🧵 RT for visibility.
In the last couple of months, I've had many conversations about using LLMs to secure the output of other LLMs.
While "inheriting" L1 security sounds like a nice idea, rollup security properties can be surprisingly tricky. Ours is the first paper to formalize these properties. If you are at CCS in Taipei, come to see our work presented! @schaliasosvons @dfirsov arxiv.org/abs/2406.16219
Excited to share that our paper has been accepted at CCS’25!
Towards a Formal Foundation for Blockchain ZK Rollups (w/ @dfirsov and @convoluted_code)
A critical question we want to answer is what it means for a rollup to be secure.
Here's what we did on formalizing ZK rollup security
Had a great time today presenting at the Research Day 2025 organized by @dba_crypto
I am personally really excited to see the tremendous progress from our friends at RiscZero. The OP Kailua numbers they cite are very encouraging, and this reported reduction in prooving times gets us significantly closer to blurring the boundary between validity and optimistic rollups, as well as realizing the vision of real-time proving.
We're planning to publicize our own numbers for the SVM shortly.
I highly recommend this great OSDI keynote from Timothy Roscoe at ETH Zurich. His key point? Modern operating systems are still stuck with Unix abstractions from decades ago. This means we're barely scratching the surface of what today's SoC hardware can do.
The same problem plagues blockchains. We've got validators running on serious hardware, but we're not pushing it nearly hard enough. Sure, Solana stepped up the game on execution speed – but there's still so much untapped potential.
At Eclipse, we're changing that.
This talk by @shemnon from EthDenver last year gives a comprehensive look at the parallelism options, beyond the usual dichotomy of declarative parallelism of Solana and the optimisic approach of Block-STM used by Aptos and OCC used by Monad.
One of the observations is that ahead-of-time access lists - both the way they’re proposed in Ethereum (TALs) and the way they’re supported in Solana can be a mixed blessing, with the burden of declaration passed onto the developer with somewhat mixed performance results compared to optimistic approaches.
The idea of delta writes is probably appropriate for contexts other than just transfers, which is a nice way to increase transaction commutativity and to reduce the de-facto contention.
Despite originating during Ethereum's Proof-of-Work era, RainBlock remains a seminal paper whose core innovations continue to influence blockchain architecture today. Some of its most enduring contributions are moving I/O operations away from the critical path, sharding the state, and its sophisticated prefetching mechanisms that minimize direct storage access. RainBlock achieves impressive results, demonstrating a remarkable 20-fold improvement in throughput. While the paper covers extensive ground, these particular optimizations stand out for their continued relevance to modern blockchain systems.
