The cheapest audit quote is usually the wrong one, this guide shows how to vet crypto, ZK, and smart-contract auditors.

Turns hash-based proof systems ZK by splitting hash-heavy work from algebra, instead of stuffing Poseidon into a circuit.

Revisits Silentium’s pseudorandom Beaver triples, adds a binary-field expansion scheme, and quietly fixes the protocol spec.

Surveys 247 LLM-agent security papers and maps prompt injection, tool hijacks, and state corruption across the agent lifecycle.

What’s the point of private inference if the agent can still be prompt-injected into exfiltrating your data?

Miden's neobank pitch: private rails, Guardian recovery, and a 3-layer stack from onboarding to reconciliation.

Paper on threshold MPC-in-the-Head signatures via Merkle commitments and TPCS; threshold sigs, now with extra plumbing.

ePrint paper on optimized ECC point-addition circuits for Shor-style DLP, with updated secp256k1 costs; hoard your curves.

A paper on oblivious garbling: the garbler only sees prescribed leakage, with linear-time malicious-secure constructions. Naturally, fewer…

Merkle trees, not giant PQ signatures, are Let’s Encrypt’s bet for post-quantum Web PKI.

AI bug-hunting for zk/MPC/threshold crypto, plus Lean4/DSL verification—because proof systems needed another adversary.

Side-by-side comparison of privacy/ZK protocols, from shielded pools to FHE, for people who enjoy taxonomy more than consensus.

SoK on PIOP-based SNARKs, with a tidy frontend/backend taxonomy and lookup/recursion fit; academia finally found plumbing.

New correlation attack on Quasi-Abelian syndrome-decoding CGs: ~1000x faster/lighter than the Asiacrypt'25 compressed-sensing jab.

PQ TLS without the handshake bloat: Let’s Encrypt wants Merkle Tree Certificates to paper over the math.

Can a mixer stay private while letting auditors watch a budgeted address set? AuditPay says yes, with ZK and selective encryption.

Paper on whether ML-KEM’s 32-byte encapsulation coin actually stays secret in practice; apparently libraries enjoy optional doom.

ZKP for full LLM inference on untrusted cloud, using sum-check and lookups to certify GPT-2/Gemma 3 token generation.

SoK on deleting public-chain data by killing keys, with a new adversary model and GDPR angle the usual immutability talk skips.

Paper on symmetric ciphers over Z/p^nZ^t: attacks, non-polynomial S-boxes, and the usual reminder that rings are annoying.

A beginner-friendly lattice crypto intro centered on ML-KEM and ML-DSA; yes, the scary math is the point.

Lagrange open-sources DeepProve, a zkML system, so you can test the “fastest” claim instead of trusting the tweet.

Why aren’t zk teams spending more on formal verification, if AI just makes specs easier and bugs louder?

Three hours down: a tweet says Zcash halted after a bad fork block number, with one dev and AI allegedly on the fix.

Ethereum's BLS exit plan: LeanSig hash signatures, zkVM aggregation, and the usual performance gymnastics.

First threshold ML-DSA: distributed signing for a few parties, with short sharing and less rejection-sampling misery.

A Zcash bug that may matter to Firo too, in a YouTube breakdown of Orchard’s privacy fallout.

What does a payments chain look like when it leans on Commonware primitives and still claims 70k TPS?

Proposes iVRFs for PoS leader election: hash-only-ish, post-quantum leaning, and less race-to-the-bottom than PoW.

Student FPGA project recreates the Bletchley Park Enigma workflow, not just the Bombe, because nostalgia needs pipelining.