Assistant to regional manager @elibensasson @StarkWareLtd
Israel
- Tweets 3,501
- Following 1,280
- Followers 5,515
- Likes 13,817
Liron Hayman ✨ retweeted
A few months later - a Quantum Safe Bitcoin tx on mainnet: mempool.space/tx/305a24ffea9…
Thanks to StarkWare's @giladi_tom85141 taking this to the finish line and @MARAFoundation_ Slipstream mining this non-standard tx
Also @robin_linus & @Ethan_Heilman on previous work led to QSB
108
155
681
253,942
Liron Hayman ✨ retweeted
If you make a NFT of a real diamond, and the diamond itself gets destroyed in a fire tomorrow, you still have the same asset.
Because the token still exists and is in limited supply just as before. Nothing has changed.
What NFT is doing to the concept of asset, few understand.
2,456
482
3,173
Bitcoin Could Be Quantum-Safe Without Protocol Changes, New Proposal Claims zerohedge.com/crypto/bitcoin…
1
6
306
Liron Hayman ✨ retweeted
Quantum-Safe Bitcoin Transactions Without Softforks
github.com/avihu28/Quantum-S…
244
667
3,405
2,013,452
Liron Hayman ✨ retweeted
Can LLMs be PROVABLE computers?
Percepta showed that a transformer can BE a computer. Compiled weights, deterministic execution, 30k tokens/sec.
But nobody asked the obvious follow-up: how do you know it computed correctly?
So I built the verification layer. A STARK that proves it 👇
1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers.
We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy
59
150
1,226
184,854
Liron Hayman ✨ retweeted
Congrats to the @Kimi_Moonshot team!
This is awesome.
Great to see this level of research coming from open-source frontier model labs.
I liked the paper so much I built a Rust implementation of it ;)
Full AttnRes + Block AttnRes with two-phase inference, built using Burn (tensor library and Deep Learning Framework, in Rust, by @Tracel_AI).
Runs on CPU, CUDA, Metal, wgpu.
Includes an interactive TUI that trains a model live and visualizes depth attention evolving from uniform to selective in real time.
Repo link and more on what is implemented in the comments.
Introducing 𝑨𝒕𝒕𝒆𝒏𝒕𝒊𝒐𝒏 𝑹𝒆𝒔𝒊𝒅𝒖𝒂𝒍𝒔: Rethinking depth-wise aggregation.
Residual connections have long relied on fixed, uniform accumulation. Inspired by the duality of time and depth, we introduce Attention Residuals, replacing standard depth-wise recurrence with learned, input-dependent attention over preceding layers.
🔹 Enables networks to selectively retrieve past representations, naturally mitigating dilution and hidden-state growth.
🔹 Introduces Block AttnRes, partitioning layers into compressed blocks to make cross-layer attention practical at scale.
🔹 Serves as an efficient drop-in replacement, demonstrating a 1.25x compute advantage with negligible (<2%) inference latency overhead.
🔹 Validated on the Kimi Linear architecture (48B total, 3B activated parameters), delivering consistent downstream performance gains.
🔗Full report:
github.com/MoonshotAI/Attent…
23
75
788
95,763
ZKML isn’t “AI on-chain” — it’s verifiable computation for for any high-cost inference / large computation.
With ZKML circuits on STWO (@StarkWareLtd ), you can prove a claim about a person (uniqueness, membership, liveness, eligibility, reputation) without revealing the biometric, the raw data, or the full output.
Think: “I’m verified” ≠ “Here’s my identity.”
On @Starknet , that becomes composable: proof → on-chain verify → permissioned access, no centralized KYC database, no silent data reuse.
Hard question: can we get to fully trustless on-chain KYC—where the verifier trusts math, not vendors?
Below Facial Recognition with Age/Gender Verification - Inference Proved On-Chain by STWO ZKML.
6
7
55
3,998
“Most altcoins are just fractionalized NFTs”
What a great framing by @joshua_j_lim on Empire
5
335
Withdrawal fee removal has changed everything
who built an automatic LP readjustment bot for Ekubo on Starknet?
Re7 Labs built one called ALMM (Automated Liquidity Market Maker).
now that withdrawal fees are gone, i want the same setup: auto reposition my ekubo LP ranges so they stay as close to spot price as possible, with minimal manual work.
5
455
Liron Hayman ✨ retweeted
It's not by chance that so many successful perps exchangesrun on @StarkWareLtd tech.
It enables running much more compute on-chain, and perps need this.
So do games and AI.
What else?
Perps on Starknet are killing it!!
Extended has now $110,084,506,509 cumulative trading volume.
record after record.
2
1
25
1,763
Liron Hayman ✨ retweeted
Starknet Trading Guild
The guild wasn’t created as a broadcast channel or marketing experiment. We wanted a place where people who actually trade could gather, compare notes, break things, and help shape how #Starknet evolves for serious DeFi users.
Our approach is simple: build in the open. That means:
1. Sharing what I’m working on as it’s being built
2. Opening up real infrastructure and real code
3. Letting this group influence what gets built next, not just comment after the fact
We believe the best trading infrastructure doesn’t come from whiteboards; it comes from traders using imperfect tools, hitting real pain points, and iterating together.
We hope to build tools that will enable members of this group to trade with each other.
Over the coming weeks, you’ll start seeing things materialize: early experiments, open-source building blocks, real strategies, and ways to actually put capital (and ideas) to work. Some of it will be rough, some will break, and some will genuinely be useful.
That’s all for now.
If you want to keep track of what we’re doing and enjoy early access, join the Trading Guild. at: t.me/+Nbf3Pst1cdhlZjM0
The first release will be in mid-February at an EthDenver event. Register now! at: luma.com/2t5kv3kn
7
6
25
4,716
