Josh Baron retweeted
For all those mystified by what the Aaronson-Ambainis conjecture really says, and what the ramifications of the proof of it really are, Scott himself has written an extremely clear explanation.
The great Scott Aaronson has written about our proof of his eponymous conjecture!
“I was surprised, honestly, to see it missing from OpenAI’s Mathocalypse dump a few weeks ago. Surely they tried it, so was it really so much harder than all the problems that were solved?”
scottaaronson.blog/?p=10295
Josh Baron retweeted
Since my post a few weeks ago on zero-knowledge proofs of ML-DSA signatures, several new works have appeared, and I also missed a few things at the time. So here’s a short update with some references.
Josh Baron retweeted
STOC 2027 received 1665 submissions. In a bold new experiment, all submissions (and author identities) are immediately made public. This is the most immediate and revealing posting of any conference I am aware of.
Take a look and see what you find!
stoc2027.sigact-ec.workers.d…
ZK-Tea: a website were verified experts in SNARKs can anonymously spill the beans on whats wrong with the benchmarks in competing papers.
(note, i'm not one of the in the weeds snark experts, I just occasionally hear comments and wish this was more public because I don't always know who to believe.)
Josh Baron retweeted
A magical corollary: EVERY FINITE SIMPLE GROUP IS GENERATED BY 2 ELEMENTS
Replying to @schweingehtbree @geoffreyirving
here we go: every finite simple group is generated by 2 elements! github.com/AxiomMath/CFSGCor…
So Anthropic has a paper on SNARKs for proving the correctness of LLM inference. The idea of snarks for LLMs is that you can check a response came from the claimed model without the operator exposing the model’s weights.
It's Cool work. TL;DR: it’s about 70-95x slower than their chosen baseline, which is very impressive.
I’m still reading, but my biggest question is whether that overhead is really ~100x. if I spent a few million tokens’ worth of engineering effort, would the baseline improve more than their SNARK scheme?
Of course, at some point this doesn’t matter. The point is cost for the capability. If you want verifiable LLM results, on a non-local model, and can’t use a TEE, is 100x overhead acceptable?
Note: they do this with a mixture-of-experts model and prove after the fact, in batches, so the prover only does work for the experts that actually ran. That doesn’t seem to translate to 100 tokens/second for other models. But like I said, I’m still reading.
anthropic.com/research/prova…
Josh Baron retweeted
🚀 @AnthropicAI's FlashProve proves AI inference while hiding model weights using zkSTARKs rooted in my papers with Alessandro Chiesa, @GiacomoFenzi & Eylon Yogev!
Never imagined my work going from pure theory to real-world blockchains to frontier AI! 🤯
anthropic.com/research/prova…
Josh Baron retweeted
Apparently 1/3 of the global Danfoss refrigerator population is flagged as honeypots by Shodan 😂
Josh Baron retweeted
Anthropic @AnthropicAI ❤️ zkSTARKs:
In collaboration with @StarkWareLtd co-founder Prof. Alessandro Chiesa and other brilliant researchers, Anthropic just open sourced a zk STARK system for proving the integrity of LLM inferencing, bringing us closer to verifiable AI.
As we said long ago (since 2018), zkSTARKs are the end-game: highest scale, best security (post-quantum). Now, also AI agile and supporting AI integrity.
anthropic.com/research/prova…
Josh Baron retweeted
The chances of being killed by a landmine delivered by a bear are low, but they are never zero.
Josh Baron retweeted
A couple of folks have pointed out a claimed asymptotically subexponential algorithm for ECDLP today. This is based on the ETH results and is very clever. But has zero practical implications whatsoever.
Josh Baron retweeted
[Revised] A gentle introduction to lattice-based cryptography (Alfred Menezes) ia.cr/2026/1098
Josh Baron retweeted
NIST has published the final version of IR 8547, its plan for retiring quantum-vulnerable public-key cryptography. RSA and elliptic-curve algorithms at every approved key size are now deprecated after 2030 and disallowed after 2035.
What changed from the 2024 draft:
- P-256, RSA-3072 and Ed25519 had no deprecation step. Now they're deprecated from January 1, 2031, like RSA-2048, and disallowed after 2035.
- A deprecated algorithm can still be used, but its owner has to examine the risk and decide. I expect auditors to ask for that decision system by system.
- A hybrid with at least one approved input, such as ML-KEM, can be FIPS 140-validated even after its classical half is disallowed.
- HQC is the alternative KEM and ML-KEM-768 the default parameter set. NIST says it will monitor developments and adjust its transition efforts as appropriate.
What NIST left open:
- Key exchange and signatures get the same dates. NIST itself says harvest now, decrypt later makes key exchange more urgent.
- In Section 3.1.2, NIST still lets authentication systems wait for a CRQC. Under Table 2, the same algorithms are disallowed after 2035 regardless.
- Contracts, registries and archives whose signatures must stay verifiable for decades get no guidance.
My full analysis: postquantum.com/security-pqc…
I cover the US requirements behind these dates in PQC Compliance: United States, a course I teach at Quantum Academy: quantumacademy.com/courses/p…
#PQC #PostQuantum #NIST #Cryptography #Cybersecurity #CISO #PKI #TLS
Imagine me telling you last year that GoAhead and eMudhra were using the same private key, with no further information. All sorts of scenarios might spin in your head, but never in a million years would you guess the real one.
Announcing 3 Critical vulns for hundreds of millions of IoT, Firewalls, OT, and Military devices including: Palo Alto Networks, Comcast, D-Link, HP, Canon, Honeywell, GE, Oracle, Siemens, Rockwell, Digi, Lockheed Martin, Boeing...
The vulns are with GoAhead Web Server which is a backend library used by these devices. There's a patch available from GoAhead but that doesn't mean there's a patch available for the devices yet -- that's up to the manufacturer.
- TLS traffic over DHE ciphers have a 1-byte secret, "04", meaning anyone can decrypt connections to these devices and steal login credentials and session tokens if the client negotiated with DHE ciphers like the latest TLS_DHE_RSA_WITH_CHACHA20_POLY1305_SHA256, which is extremely likely.
- LibreSSL mTLS has been completely bypassed, meaning anyone can connect over mTLS. If you have one of these devices listening on the internet and are expecting mTLS to protect it from attackers, take it offline or patch asap.
- OpenSSL Certificate Revocation Lists (CRLs) are not enforced, meaning if a user device is stolen or compromised, and you revoke that mTLS cert, nothing actually happens. There are a lot of other attacks this can be used for.
These are critical because these types of devices run our world. Water systems, electricity, military, firewalls, critical infrastructure.
foxio.io/blog/choose-your-ba…
Josh Baron retweeted
NIST’s new PQC migration guidelines can help organizations transition to post-quantum cryptography standards. This report outlines the timelines and activities for federal agencies and serves as a resource for organizations navigating the transition.
🔗 nist.gov/news-events/news/20…

ALT A digital illustration shows a transition from old encryption standards to new encryption standards. On the left, an outdated server is connected to symbols and codes. In the center, a large, futuristic quantum computer is depicted. On the right, a shield represents the new encryption standards, with symbols indicating modern data security. The background features a binary code theme.
Josh Baron retweeted
The largest math proof in human history is now computer-verified.
Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry."
About 15,000 pages, built by over 100 mathematicians across generations.
Josh Baron retweeted
Regarding the Exponential-Time Hypothesis, this is what I wrote a decade ago!
cacm.acm.org/opinion/the-mor…




























