Fast walker, slow talker. Maintainer of binius.xyz. Prev CTO @IrreducibleHW.
Berlin
- Tweets 386
- Following 417
- Followers 2,938
- Likes 2,146
Irreducible's time has come to an end.
Long story short, @radi_cojbasic and I ultimately came to the realization that we couldn't sustainably build the kind of deep tech business in ZK that motivated us.
46
15
247
64,500
Breakdown of different types of Bitcoin clients medium.com/@jimpo/flavors-of…
2
47
138
By the end of grad school in 2020, I was convinced ZK was way too slow and difficult to be useful.
Plonky2 and ZK hardware changed my mind.
Now with Binius, we'll show how wrong I was then.
4
13
96
9,105
Very excited to open-source PetraVM and build in public! Petra is how we will unlock real-time proving for arbitrary computation and limitless blockchain scaling.
WebAssembly -> Petra -> Binius -> Irreducible's HW
Can't keep up with all the new zkVM announcements? Here's one you don't wont to miss.
Today we are introducing a preview of PetraVM, a Binius-based verifiable VM built in collaboration with @0xPolygon!
7
16
88
12,461
I'm super proud of the whole Irreducible team for pulling this together! We finally have a live production service and an amazing and fast developer platform for ZK. binius.xyz/
We're about to take Binius to new heights, now's the time to get involved!
🚀Big news! Irreducible launches alpha-ready Binius library and its first application, an Ethereum state proving service. Here’s what it means: 🧵
3
6
85
8,212
Feels great to finally announce what we've been working on at Irreducible the last few months!
Binius64 is in many ways what Binius V0 should have been, but we didn't know that at the time.
1/ Today marks the beginning of a new chapter for Irreducible. We started as a custom-hardware company, and now we are fully focused on high-performance software.
Read below about our pivot and our new proof system, Binius64 👇
9
12
78
11,616
Pet peeve: when people say DDoS when they really mean DoS. Instant loss of credibility.
11
1
70
17,832
A summary of some ideas on Bitcoin commitments. medium.com/@jimpo/next-gener…
2
16
65
Key insights for blockchain design:
1) Optimize common-case settlement.
2) Make common-case settlement of different contract types indistinguishable.
3) Use primitives secure under well-understood assumptions.
4) Schnorr does all of the above!
Video from @pwuille's talk on Taproot, Schnorr and SIGHASH_NOINPUT is now available piped.video/YSUVRj8iznU
5
18
57
The Binius64 codebase has been relicensed to MIT + Apache 2.0, and I plan to spend some effort maintaining and improving Binius64... at least for now.
1
1
50
2,766
Thrilled to be working with @BruestleJeremy and the amazing RISC Zero team on this!
RISC-V is a perfect use case for Binius arithmetization using tiny, power-of-two-sized data types with no embedding overhead.
Excited to announce our partnership with @RISCZero to integrate Binius, bringing ultra-fast performance to the original RISC-V zkVM.
Together, we're setting new standards for ZK performance and accessibility.
irreducible.com/posts/irredu…
3
1
47
4,632
I'm proud of Irreducible's innovations with Binius, and I hope this work has an impact. I still believe firmly in the core Binius techniques for ZKPs: transparent hash-based, multilinear IOPS over tiny binary fields.
1
1
46
3,296
With BIP 157, unless you are eclipsed or the longest chain is invalid, then you are 100% safe. If eclipsed by a low hash-rate adversary, worst case is you miss some txs or download a little extra data. If you think all light clients are inherently evil, I doubt I'll convince you.
5
42
Calling all reviewers who would like to see Core support for Neutrino-style light clients.
Step 2/3 for getting neutrino into Bitcoin Core by the amazing @jimpo_potamus github.com/bitcoin/bitcoin/p…
1
7
36
It's an honor to have met and collaborated with so many talented people in the ZK space, and learned so much from my brilliant colleagues, especially @benediamond and @lockedloop.
3
2
40
4,854
I'm so grateful for the opportunity, the support from our awesome investors, and the crash-course in Serbian culture from @radi_cojbasic. :-)
Hell of a ride
1
39
3,944
🚨I’m looking for an Engineering Manager to partner closely with me and lead our exceptional Cryptography Engineering team. You'll own team productivity, engagement, and happiness, allowing me to focus more deeply on technical direction.
3
5
30
5,776
Giving a talk Thursday on the Lightning protocol. LA homies, come out in droves! meetup.com/LA-Blockchain-Dev…
2
10
28
Last night I tried buying ice cream with Lightning and the payment got stuck pending. It is still pending 13 hours later. That really happened!
As an engineer who used to warn about this, it's kind of interesting. As a customer who wants ice cream, it's a terrible UX.
6
23
Excited to announce that I'll be a speaker at the StarkWare Scholar Summit in NYC on August 5!
Join us here: lu.ma/n9yrzbvf
2
1
20
670
Why Berlin? Everyone keeps asking.
This is why @rcojbasic and I decided Berlin is the *best city in the world* to build Irreducible. 🧵
4
4
23
1,770
Announced today what we're building @UlvetannaHQ!
Lots of fun problems at the intersection of high performance computing and applied cryptography -- hit me up if you're interested in working together!
ulvetanna.io/news/introducin…
1
22
2,051
This project was only possible thanks to how awesome the team at Polygon is, led by @RobinSalen, and @bfarmer and @dlubarov 's belief in the binary field vision
3
21
4,136
We are actively developing the open-source Binius implementation here: gitlab.com/UlvetannaOSS/bini…
We take a lot of inspiration from plonky2 & Plonky3, and even use some Plonky3 crates. As soon as benchmarks are ready for hashers and u32 add & mul, we'll publish the comparison!
1
19
1,350
I'm so happy to see community contributions to Binius picking up. This week, the Binius project merged PRs from three external contributors!!! 🤩
Thanks @tcoratger, TomWambsgans, and storojs72, you guys rock!
17
793
I’m seeking someone endlessly curious, relentlessly focused, passionate about organizational health, and excited to work in-person alongside @radi_cojbasic and me in Berlin.
If this sounds like you—or someone you know—DM me directly or apply here jobs.ashbyhq.com/irreducible…
1
8
576
I found an optimization to Caulk (by @arantxazapico & co) while working with Akis Kattis and Ulvetanna!
Now the lookup argument has prover complexity independent of the table size (with precomputation)
eprint.iacr.org/2022/957
1
2
19
Great work by the @RiscZero team on making it as easy and cheap as possible to prove computation, leveraging economies of scale!
I love the summary diagram on the last page.
Introducing The Boundless Protocol
Designed to scale ZK-powered verifiable compute for every chain. Making it more efficient, reliable and accessible.
3
1
17
1,278
Learn to think in M3! Here's a short & fun talk I gave recently, involving mazes, virtual machines, and Colored Petri Nets
Link to the recording:
piped.video/3ua-t9sDyes
Thanks to all the sponsors of ZK Hack Meetup in Berlin:
@IrreducibleHW
@powdr_labs
@anoma
💜💜💜💜
14
965
Always a joy talking with @sinahab!
Sina is an amazing, curious, and very thoughtful friend and I'm honored to get to be on his podcast, Into the Bytecode, talking about Binius and @IrreducibleHW.
Here is my conversation with @jimpo_potamus, cofounder/cto of @IrreducibleHW.
Jim is a brilliant human being - he was one of the first engineers at Coinbase, went back to academia for a grad cryptography degree, is now making core contributions to zk.
Here we talk about:
- Binius as a proof system using binary fields
- FPGAs, hardware, running a data center
- the verifiable internet
Links below.
1
1
16
1,090
@gakonst was one of the first in the industry to understand and articulate the importance of ZK hardware, and is a big part of the reason I'm working on it today
I wrote about ZK Hardware 2 years ago, and their eventual rise. Really exciting to see this play out.
It's totally possible there are proof pipeline internals that we can freeze and build ASICs for today, e.g. the proof composition step to MSM snark a la Groth16/FFlonk before pushing the proof on-chain since the chain also change much.
But it still seems hard to think we've ossified earlier in the pipeline, esp. given the 12+ mo tapeout time.
paradigm.xyz/2022/04/zk-hard…
1
15
4,099
If anyone is interested in using and/or further developing Binius, I'd be happy to talk to them. Or hiring any ex-Irreducible folks looking for jobs. They're all very smart!
1
2
15
2,563
Discussing the state of blockchains with @zmanian, @iam_preethi, and yours truly
2
11
Very impressive results!
Exciting to see more teams realizing the advantages of multilinear proof systems and bridging the gap between beautiful theoretical results (where @ronrothblum has been a pioneer) and concrete performance
3/ With SP1 Hypercube, we combined novel cryptography and relentless performance engineering to become the first zkVM to achieve real-time Ethereum proving.
In a benchmark of 10,000 mainnet blocks, 93%+ of blocks were proven in <12 secs, with an average proving time of just 10.3 secs.
7
945
Super impressive! Kudos to the Polygon Zero team for pushing the limits!
We've got more work to do on Binius perf to keep up
After some optimization work at @0xPolygonZero, Plonky3 is up to ~2,500 Keccak-f permutations per second on an M3 Max! In other words, we can prove one of the least ZK-friendly hashes at a rate of about 340 KB/s.
11
1,395
Our development velocity shot up and the entire team could work together more cohesively. We've forced ourselves to embrace a mindset of shipping faster, ruthlessly cutting scope, and making the best possible use of modern AI.
1
1
13
622
We have so many more ideas for directions to take this, but want to get user feedback to guide us now.
I'm extremely proud of the whole @IrreducibleHW team for pushing the state of the art. We're not stopping until the Internet runs on ZK.
11
567
Please do! Code's still active development, but feedback is welcome.
gitlab.com/UlvetannaOSS/bini…
2
2
11
1,245
It's not for 17, but it doesn't matter because two more PRs are needed before it's useful. Coming soon to a GitHub repo near you...
1
10
An overview of BOLT Lightning from the awesome LA Blockchain Developer's meetup last month piped.video/Ysj2yobFMF4
1
2
10
For @radi_cojbasic and me, putting aside the FPGAs and all the code that we'd spend the last 18 months on was an incredibly hard call. But once we started on Binius64, we knew it was the right one. The second time around, we knew what to build and how to build it right.
3
10
1,685
John Oliver says crypto bakers will be gainfully employed in 50 years. He's clearly pretty bullish on Tezos. piped.video/watch?t=12m2s&v=…
1
8
Well, if you search me on Google, this is the search blurb
There's literally nothing more insulting than being wrongly associated with Ada and EOS
9
Yeah, custom ASIPs (app-specific processor) will be best eventually, but FPGAs aren't enough. Hope ZKP adoption picks up enough for that to become reality. Nvidia just adding carryless mul instructions would go a long way...
1
1
90
Everyone knows that "ZK" stopped meaning ZK a while ago. Until this summer, Irreducible was building high-scale verifiable computing but ubiquitous true ZK was the end goal. We hoped to get there with custom silicon, but the timeline was too long and today's market too small.
1
8
285
The competition in the ZK space is fierce and we're up against some of the smartest teams out there. This spring's announcements of GPU-accelerated real-time Ethereum proving was a wakeup call. And on the other side, Google's ZK-Longfellow system lit a fire under our asses.
1
8
261
Real talk, my desktop background is an NFT that I don't own and I haven't lost one second of sleep over this
2
8
I'm incredibly proud to announce I just re-wicked a candle with a match and a coffee stirrer stick.
Friday night is off the chain.
7
The code is still public for anyone who wants to resurrect this
github.com/jimpo/march-madne…
I was an oracle for one in 2017. May have been the first bracket on Ethereum 🏀
teddit.net/r/ethereum/commen…
1
1
6
1,903
This year, Irreducible will ship Binius into production. We're collaborating with industry-leading teams like RISC Zero and Polygon to develop groundbreaking zkVMs. Powered by our FPGA clusters, they will make verifiable computing practical for tons of new use cases.
1
1
6
675
"A Practical Guide to Linux Commands, Editors, and Shell Programming" - Mark Sobell
That really cemented my interest in being a programmer.
1
7
I've started calling everything that uses linear codes & hash functions "STARK-ish" protocols (like using FRI or tensor IOPs)
2
6
583
The client should be open-source, not the server. The idea of E2E encryption is that the backend can be as malicious as possible and privacy is still preserved.
1
6
Come help us build world-changing technology and an awesome in-person culture in Berlin! We're hiring!
6
296
First, Electrum does get PoW from all connected servers, then chooses one of those at random to sync txs from. Second, even if you do something similar with BIP 157 (trust a random node for correct filters), privacy is far better than with Electrum
1
5
The house I've been living in for 6 months has a *_5G network which is 2.4GHz and a *_2G which is 5GHz. Today I finally swapped them and now my world makes just a little more sense.
6
Big move by @SocketSecurity! They acquired Coana to integrate their static analysis engine and reduce false positive alerts for code vulnerabilities. The progress that Socket is making on securing software from open source suppy chain vulnerabilities is super impressive!
We just bought a company.
Why? Because vulnerability scanning is fundamentally broken. And I’m tired of pretending it’s fine.
We acquired Coana, the best reachability analysis engine on the planet.
The whole vuln industry is addicted to quantity over quality. More alerts, more dashboards. It’s security theater. And it doesn’t scale.
Ask any developer what they do with 1,000 security alerts. They ignore them.
Coana flips the script. It asks the only question that actually matters: Is this vulnerability even reachable by your code?
If not, who cares? Move on.
If yes, fix it fast. 💥
I’ve been obsessed with this idea for years. Why drown developers in noise when we can tell them what actually matters?
When I first saw Coana, I knew: “We have to bring this into Socket. Nothing else even comes close.”
Coana doesn’t just work — it’s freakishly good.
✅ 80%+ fewer false positives
✅ Instant results with first-of-its-kind "precomputed reachability"
✅ Full source code access is optional
✅ Can even run offline on air-gapped networks
Yeah. It’s that good.
I’m thrilled to welcome @ndrssndrgrd, @torp_martin, @amoellercsaudk, Benjamin, and the entire Coana team to Socket!
These are world-class engineers. Real researchers. Together, we’re going to build security that actually scales — and actually helps you ship faster.
Legacy tools collapsed under modern dependency trees. We’re not here to patch the current system. We’re here to replace it.
This is how we move the industry forward. 🚀
1
1
4
1,473
The input is a set of trie key + value blob pairs and it verifies inclusion in Ethereum's Merkle-Patricia Trie (Keccak-256) with respect to a block state root.
Limitations:
- No storage trie proofs yet
- No account trie update/insert/delete handling yet
1
1
5
901
According to FGPA devs at @UlvetannaHQ:
Latency is ~8x better for goldilocks and resource utilization is 10-15x better, meaning you can fit that many more onto one board
5
220
For those in the US unhappy with both political parties and your presidential candidates: there is a way out and it's called *ranked choice voting*
RCV -> viability of more parties -> parties that represent your interests -> elected officials with sensible, not fringe positions
#RankedChoiceVoting is the fastest-growing election reform in the country. Check out our legislation tracker to see where RCV is headed next! fairvote.org/ranked-choice-v…
1
5
706
Thanks for the shout out! Very much still a WIP. Hopefully will have some interesting results to share when it's ready.
4
"Crucially, blockchain creates a record which is incorruptible."
Great piece, Eli! This is why we build blockchains.
Voting's on everyone's mind lately.
My 2 cents: move'em to blockchain.
eureporter.co/politics/democ…
1
5
1,099
# of users who use trusted 3rd party nodes
Apps that recommend trusted 3rd party nodes
Otherwise interesting projects that require worthless tokens
History of setting hilariously unrealistic expectations
*All the complexity*
5
opensea.io/assets/0x495f9472…
@JeremyRubin did! It's a once in a lifetime opportunity to pick up your very own (NFT of a) iconic bridge.
2
4
The account has been hacked, this is a scam.
We're working on regaining access. @Support please help.
2
4
268
Verifying myself: I am jimpo on Keybase.io. CtR6tykL3aox96vS_rwt4FbwKl0-qOuK6GAW / keybase.io/jimpo/sigs/CtR6ty…
5
Top 5 qualities they should have:
1. Integrity
2. Good judge of character
3. Handles stress well
4. Raw intelligence
5. Knowledge of history
1
1
4
We have a formalization of this in our paper "Proximity Testing with Logarithmic Randomness" (eprint.iacr.org/2023/630). See definitions 4.3 and 4.4. The idea of the admissible query sampler came from Marlin.
1
4
232
Size of the circuit doesn't matter as much as the nature of the computation being proven/verified. As @bfarmer said, Binius will shine for hash functions involving many bitwise operations and is somewhat disadvantaged for integer multiplication. We think the advantages outweigh.
2
141
Wrote some infrastructure/tooling for managing my Ethereum validator
github.com/jimpo/eth-staking
1
4
Got nerd-sniped hard.
Came up with a signature scheme that could be used for this sort of situation. They're not quite ring signatures, and they're more than randomizable signatures. They're ringimizable signatures! Is this already a thing?
hackmd.io/eNa1xld_SaCLY1S0xq…
Very interesting design space!
You could also prove an art piece was created by one of several famous artists without revealing which one
4
The prover costs for the polynomial commitment and sumcheck protocols are in the performance section. Commitment costs are quasilinear in the number of bits committed (due to RS encoding), regardless of field size committed.
Unaware of that work -- we'll review and follow up
1
4
254
It'd use the index interface, not the P2P one that SPVs do.
The rescan logic in the codebase (github.com/bitcoin/bitcoin/b…) reads every block in the rescan range from disk, parses it, and iterates through all txs. Instead, it could read from this filter index first, which is cheaper.
4
Maybe soon! We've been researching this recently (see eprint.iacr.org/2025/1015)
What are the ZK applications you are most interested in?
2
67
Right, in the absence of a reliable PKI you need to verify safety numbers to defend against MITM. Server source is better than nothing, but still not a full solution.
4
Demand real democracy reform or we won't break the cycle.
Demand proportional allocation of electoral votes (or election by popular vote). Demand campaign finance reform and transparency. Demand voting rights for all citizens, fair districts, ranked choice voting.
4
There's Shoups lower bound proof for generic groups. shoup.net/papers/dlbounds1.p…
And I know discrete log is easy in characteristic-2 fields.
1
4
"That caused a brief interruption in the [Daily Stormer's] operations but they quickly came back online using a Cloudflare competitor. ... I have little doubt we'll see the same happen with 8chan."
It's almost like you didn't read the post or don't care about facts.
1
4