🧵
We updated ePrint 2026/1692 with new results on Poseidon: eprint.iacr.org/2026/1692

The paper introduces a new approach to round skipping which we denote (k+1)-wise collapsing and which enables round skipping in arbitrary CICO-k problems beyond CICO-1 and CICO-2.
1/6

DeFi meets research, industry, and fresh ideas. 👀

The upcoming MidChain DeFi Conference, organized by @MasterkeyVC, is bringing the worlds together to talk about what’s next.
@tzahi_kanza and @AvriRotem from Syndika will be there - come say hi!

Thanks to @MasterkeyVC, Devs and @3miLabs for making it happen. 🚀

Research into applied cryptography is generally spread across universities and not always connected
to real systems. @3miLabs, founded by cryptographer @TomerAshur and associated with @KU_Leuven's
@CosicBe lab, addresses this by bringing researchers directly into client teams.
[...]
“Privacy and security are not optional in systems that hold real value. When cryptography fails, it
usually fails silently and irreversibly. Only a few hundred researchers worldwide can design these
schemes and prove they hold, and 3MI Labs brings a group of them under one roof,” adds Jakob
Lingg, partner at Vsquared Ventures.

There have been some conversations over the last couple of days on what kind of cryptography is going to be hit first from AI improvements. I don't know enough about lattices and other PKC hardness assumptions, so here's my two cents as a symmetric-key cryptographer with above average experience in hash functions (both traditional and algebraic) >>

In essence, the adversary skips the first two full rounds and tolerates a degree growth in a third full round before proceeding to skip partial rounds in a non-linear GSR. This forms a low degree ideal that can then be solved using Gröbner basis methods.
5/6

3MI Labs continues to evaluate the security of Poseidon from first principles as requested by the EF via grant number FY26-2457.

Shoutout to our researchers @kr_ams, @sundasguria1, and @TomerAshur for this result.
6/6

RT @alignedlayer: A proof convinces a verifier.
A spec convinces a person.

Lambda VM gives you both. ~60k lines of Rust, one chip per file…

[Revised] From Toy to Instrument: Seven Years of Verifpal (Nadim Kobeissi) ia.cr/2026/1654

POV: arriving in the office every morning

I did not yet have the time to read this paper property but if the results as claimed are correct this is very impressive and I think that the EF was wrong in dismissing it as out of "intended" scope.

Smth cookin'

RT @alignedlayer: ai can write more code in a day than anyone can audit in a year.
lambdavm went the other way: minimal rust, few dependenc…

Our team's latest paper, "Experimental Evaluation of Dickson Polynomials as an S-Box for Algebraic Hash Functions", has been officially published in Intelligent Computing (LNNS, vol. 2064, Springer): link.springer.com/chapter/10.100…

The key finding: Unexpectedly, experimental results show that the solving degree of the resulting multivariate polynomial system remains the same, despite an increase in time and memory overhead.

📖 Read the paper here: link.springer.com/chapter/10.100…