drouyang (@drouyang) profile picture

co-founder & ceo @snarkify_zkp | phd in operating systems | ex @Meta

Joined August 2010
Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)Photo by drouyang (@drouyang)
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
2,365
7,275
5,745
55,949
9,967,793
drouyang retweeted
Hard to overstate how big of a deal this is. tl;dr — ANY unpatched Linux server with even a single TCP port exposed to the internet (say for HTTP or SSH) and MPTCP active (default for many Go web servers) is vulnerable to complete takeover.
196
659
123
10,189
1,496,696
drouyang retweeted
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa
4,171
7,148
4,761
69,063
13,738,572
drouyang retweeted
.@apolynya is back with updated thoughts on ZK. One line stood out: “The potential impact of ZKPs supercedes that of all blockchains.” Early signs say he’s right. Our upcoming 2026 ZK Industry Report with @zeroknowledgefm surveys adoption across categories beyond blockchains.
9
16
1
88
10,626
Introducing Magnetar from Microsoft Research, the first zkVM that is: (1) general-purpose (RV64), (2) post-quantum from standard hashes alone (no curves, no lattices, no Poseidon), and (3) built for recursion. Despite its conservative security, it is the fastest zkVM: 7.2–8.7 MHz on a laptop, 10–39× faster than existing zkVMs across 15 workloads! An optimized version reaches 9.5–12.8 MHz. Links in the thread below.
54
94
43
616
63,944
1/ Today we're releasing Lattice Jolt: a post-quantum version of the Jolt zkVM, built on lattices instead of elliptic curves. It's _faster_ than curve-based Jolt and has the shortest proofs of any post-quantum zkVM — under 100 KB. Post: a16zcrypto.com/posts/article…
30
135
32
736
101,351
drouyang retweeted
We believe developers should choose their own security configurations, but we made a mistake by allowing our DVN to act as a 1/1 DVN for high-value transactions. We didn't police what our DVN was securing, which created a risk we simply didn't see. We own that. Moving forward, LayerZero Labs will be more active in educating developers and monitoring how applications should build on the protocol to ensure they are configured safely.
21
9
1
92
7,956
drouyang retweeted
Local coding models may have crossed a threshold to be viable. We compared several 30B class open weight models on Terminal-Bench 2.0 and they are close to what frontier models were around ~6 months ago. - For consumer hardware: MoE > dense when memory is the bottleneck. - We expect this can be improved a lot with some prompt/harness/llama.cpp tuning
2
2
1
16
122,341
Zero introduces four 100x breakthroughs across storage (QMDB), compute (FAFO), networking (SVID), and zk proving (Jolt Pro). It lives up to everything we stand for: - Decentralized - Permissionless - Censorship-resistant
17
43
18
396
217,098
drouyang retweeted
Snarkify wrote an SP1 GPU prover from SCRATCH to brrrrrrr on real-time proving. Can't wait to see their results on the Succinct Prover Network.
After 6+ months of engineering, the @Snarkify_ZKP team is nearing the launch of SP1-SRT — a fully rewritten, GPU-accelerated SP1 prover for distributed real-time proving. It’s fast. It’s cost-efficient. And it’s coming to Succinct Network. Get ready.
192
64
6
523
17,901
Excited to welcome Snarkify, an elite hardware acceleration team and serial ZPrize winner, to the Succinct Prover Network. Their expertise in cryptography, high-performance computing, and cloud operations will be valuable to our customers by helping reduce cost and latency.
Snarkify has arrived on the Succinct Prover Network.
49
12
3
109
5,374
🤯 real-time proving is here 🤯 Mainnet EVM blocks proven in under 1 Ethereum slot (12s). Goosebumps. Succinct proves every Ethereum L1 block: → 94% in <12s → 99% in <13s → 99.9% in <12s, soon™ Yesterday RISC Zero unveiled a $120K home GPU cluster—proofs expected in 9.25s. Brevis, OpenVM, Snarkify, ZisK, ZKM are weeks from joining the real-time club. Soon™ my validator will verify EVM blocks on a Rasberry Pi Pico—a $5 board that consumes <1W. I will ditch my EL client in favour of a zkEL. No 1 TB NVMe. Goodbye Geth, hello zkReth. Stateless and RAMless verification in milliseconds on a single CPU core. With real-time proving 1 gigagas/sec (10K TPS) is within reach, without compromising validator decentralisation. From now on expect regular gas limit bumps. 10% of stake is already voting for a 60M limit—your validators can too. Snarkifying mainnet turns Ethereum L1 into the first based and native rollup. Stage 2. Bug-free. Decentralised sequencing. No security council. No governance. The L1 will lead by example. This Friday we celebrate. Join us for Ethproofs call #2, May 23 at 2pm UTC. 25 speakers, 2 hours of content. Calls are open—DM @corcoranwill for a calendar invite. We are witnessing history. Believe in something real. Believe in real-time proving.
224
433
126
2,469
400,308
Very happy to see Tower Research, an HFT powerhouse, putting a solid piece on ZK proving.
1
9
680
Here's my stance :) We can scale the canonical L1 EVM by 1,000x—from 10 TPS to 10K TPS—using zkVMs. The core idea: validators verify a SNARK for every EVM block, instead of re-executing it. I'll discuss this on @eth_proofs call #1, Fri Apr 25 at 2pm UTC. The founders of 17 zkVMs snarkifying the EVM will share updates. If there's one white-hot corner of crypto right now, it's zkVMs. The pace of progress is mind-blowing. DM me or @corcoranwill for a calendar invite to Ethproofs call #1. That said, most users will still live on L2s. Why? Because L2s will offer a combined 10M TPS—100x from rollups, 100x from Danksharding, 100x from a decade of Nielsen's law. So the L1 EVM only gets 10K / 10M = 0.1% of activity, and that's perfectly fine. The key to graceful horizontal scaling is preserving Ethereum-wide network effects: → native rollups = shared security → based rollups = shared composability → pristine ETH = shared money Ethereum's future is incredibly bright. Believe in something :)
77
142
64
1,029
257,113
Can’t be April fooled when secured with ZK. Can be fooled any month when secured with trust.
2
1
10
441
The @Zk_Hub_Global presentation by Dr. Jiannan Ouyang, CEO of @Snarkify_ZKP, is now available on the @HouseofZK YouTube: youtube.com/watch?v=7Iw_2cnX… @drouyang’s talk focused on GPU acceleration in zkVMs and the challenges of achieving real-time proving. He outlined the evolution of GPU optimization, from early kernel-specific enhancements (e.g. MSM, NTT) to end-to-end zkVM acceleration, emphasizing that GPUs outperform ASICs for ZK due to memory constraints. Snarkify aims to achieve real-time zkVM proving by June 2025, which could be a significant factor in driving mass ZK adoption 🤝
11
8
1
39
1,894