Andrew Tretyakov
@0xAndoroid
Followers
331
Following
161
Media
15
Statuses
170
Engineering at @a16zcrypto — building Jolt zkVM; Math'26 @UWaterloo; Rustacean 🦀
Joined January 2022
First time paying with USDC. No gas fees, no KYC. Just scan the QR code and click a button in a wallet. I'm actually surprised as to how convenient it is, literary faster than typing in credit card numbers. The only problem is that Base is the only supported network.
1
0
1
Worked so hard that I completely missed ETH all time high. Congrats to everyone in the trenches.
0
0
3
RISC-V is the best target for Jolt. All RISC-V instructions are very sumcheck friendly. You can prove correctness of instruction output in small memory, without materializing entire 2^128 table.
1/ Should Ethereum enshrine RISC-V at Layer 1?. This was the central topic of the recent Ethereum Foundation proofs call #4. The discussion touched on zkVM design, efficiency tradeoffs, and Ethereum’s long-term execution model. Let’s break it down 👇.
1
0
4
Literary crushing any other zkVM.
1/ Jolt just got a 6× speedup. We now prove over 1,000,000 RISC-V cycles/sec on a 32-core CPU and over 500,000 cycles/sec on a MacBook, with proofs of about 50 KB. Here’s what’s new 🧵.
1
2
24
Just run some random linux voodoo commands as root, straight from GPT, and Jolt became 40% faster.
0
0
2
They never shipped it 😭😭😭.
0
0
4
RT @NoahCitron: Helios 0.9.0 is now available! This release is packed with new features and improvements:. • A functional TUI.• Historical….
0
21
0
Surprised I still need to do approvals to swap on Ethereum given that #EIP7702 went in a couple of months ago.
0
0
1
After touching most of Jolt's codebase over two weeks, the architecture finally made sense: ~90% sumcheck protocol. Whether it's Twist, Shout, or Spartan gluing everything together - the mathematical beauty is how sumcheck elegantly powers it all. It’s just so simple.
0
0
4
The biggest realization I got in the first week working on Jolt is that all zkVMs (SP1, Risc0, Jolt, etc.) are not actually ZK, they are just succinctVM. Unless you wrap the proof in a Groth16, there are no hiding guarantees.
0
0
9