I know it’s not most folks‘ primary concern, but LLMs or not, I’m unimpressed by how soft these infrastructure services are. What do you mean HF had a Jinja2 template injection. And I’m still not over GitHub’s unsandboxed RCE. Geomys might need to self-host code/CI to avoid a weak link.

I’m not sure I like this new world where we write software for other software to use. What’s even the point of it?

I have gone very far into the weeds in investigating every (I think) published mechanism for preventing authentication tokens from being stolen and how basically none of them has actually succeeded in a useful way: www.codon.org.uk/~mjg59/blog/...

Bernstein's main achievements, by decade: 1990s: Bernstein v. United States 2000s: ChaCha20, Poly1305, Curve25519, Ed25519 2010s: SPHINCS 2020s: sabotaging the post-quantum transition This is so f***ing frustrating. And sad. But mostly frustrating.

me: isogenies are beautiful and elegant and allow small sizes also me: vroom vroom lattice go vroom

Deirdre Connolly¹ ²@durumcrustulum.com · last mo.

lattices are so fucking fast 🥹

i’m going to make a fake pcpartpicker website where you can shop a whole build using 2019-like prices and you don’t actually pay money or get a computer but at least everyone has a little fun

There we go. US Gov tightens post-quantum cryptography transition deadlines for high-value systems to 2030 for key exchange and 2031 for signatures. Also, speeding up the CMVP (FIPS 140 validation) processes. That’s how you know the rush is real. The quantum computers are (potentially) coming.

Securing the Nation Against Advanced Cryptographic Attacks

By the authority vested in me as President by the Constitution and the laws of the United States of America, it is hereby ordered: Section 1.  Background

whitehouse.gov

i'm a software engineer and once got budget from my director to hire a writing instructor for the department. cost was peanuts to us (like $5k?) but the benefits were enormous. she wound up sending all the managers through it too. massive low hanging fruit for any eng director who wants it

There would have been responsible ways to write that rule, by the way. For example, by making it turn itself off after a few months if not updated. Palo Alto has no incentive not to harm the Internet or their customers though, because IDS/IPS are usually mandated.

In 2020, OpenSSL had a vuln related to the signature_algorithms_cert ext. openssl-library.org/news/secadv/... Palo Alto apparently "solved" this in their IPS by blocking connections with "unknown" algs in sigalg_cert. Six years later, we can't add ML-DSA to sigalg_cert in Go. sigalg_cert is dead.

crypto/tls: TLS 1.3 handshake timeout with tip, Go 1.26 can connect successfully · Issue #79626 · golang/go

What version of Go are you using (go version)? $ go version go version go1.27-devel_bfbbe9667e Fri May 22 15:49:08 2026 -0700 darwin/arm64 Does this issue reproduce with the latest release? No What...

github.com

I wrote a new post about the privacy risks of on-phone agents like Apple’s new Siri, and how private inference isn’t any sort of silver bullet. blog.cryptographyengineering.com/2026/06/09/a...

The future of Siri, or: why private inference isn’t private enough

Yesterday Apple announced a big step towards deploying real AI in their Siri ecosystem. In most ways this is good and inevitable: Siri is one of the world’s most widely-used voice agents, and…

blog.cryptographyengineering.com

Modern technology could learn from older stuff that does what it's supposed to and doesn't bother you. My toilet just works and never sends me emails like "Good news! Flush® is now Floosh®"

I mean "why is there a special IPSec AEAD in the kernel wired up to splice() and why is that enabled by default in everyone's kernel" complexity.

super excited for the day AI will be able to take over the most tedious tasks of daily life such as making art, engaging with art, maintaining human relationships, etc., so that I can focus on my true passion: creating value for shareholders

yesterday I reported on GitHub employee concerns about reliability and leadership, and then hours later GitHub suffered a catastrophic outage 😬 www.theverge.com/news/918001/...

listen hun if you can't handle the Washington Post editorial board at their the Texas gerrymander "freakout" is "not a threat to democracy" you don't deserve them at their "Virginia plunges America deeper into the gerrymandering abyss"

Palantir are about six months away from ordering their employees to leave audio logs scattered around their offices

It's April 2026, 1 year 8 months since FIPS 204. The IETF TLS WG is busy debating the concept of ML-DSA hybrids, and whether they should be composite, concatenated, or separate. The complexity of hybrid auth is, however, firmly denied. In the distance, sounds of a pure ML-DSA PKI being built.

☂ internetfriends.llc

Internet Friends LLC

A software company in California.

internetfriends.llc

Setting up SAML between an app and an identity provider should be two clicks. Instead, it's 20 manual steps of copying and pasting and clicking, as explained in a bespoke PDF for each app. With multiple failure modes that will lock you out. Huge failure of protocol design.

I have been saying for a while that given use of unsafe languages and terrible dep hygiene you’d think things would be much worse than they are. With better LLMs and more determined attackers like in the Trivy and Axios attacks, I’ve stopped saying that. Worse arrived slowly, then all at once.

The idea that big tech companies will inherently have better security tools than me because of enterprise access deals makes me sad.

🚨New preprint and our results are rather concerning.. We find the "boiling frog" equivalent of AI use. Using large-scale RCTs, we provide *casual* evidence that AI assistance reduces persistence and hurts independent performance. And these effects emerge after just 10–15 minutes of AI use! 1/

We should be able to slow down AI takeover by a few years just by telling the model to find every bug in ffmpeg.

“Everyone from Walmart to your local bodega will need to maintain a presence in Fortnite to stay relevant” is so profoundly stupid I’m just glad I have the documentation to prove someone legitimately believed it.

Two papers came out last week that suggest classical asymmetric cryptography might indeed be broken by quantum computers in just a few years. That means we need to ship post-quantum crypto now, with the tools we have: ML-KEM and ML-DSA. I didn't think PQ auth was so urgent until recently.

A Cryptography Engineer’s Perspective on Quantum Computing Timelines

The risk that cryptographically-relevant quantum computers materialize within the next few years is now high enough to be dispositive, unfortunately.

words.filippo.io

On the topic of tokens, it is very wrong that a project on GitHub or NPM can insist on 2FA for people logging in, but then those same systems allow using these short easily stolen strings as 1FA methods with equivalent power. Recent attacks demonstrate the significant lateral movement this enables.

Dan Lorenc is exactly right: "I blame[] the Trivy breach on GitHub. The design of Actions is plain irresponsible today and ignores a decade of supply chain security work from other ecosystems." www.linkedin.com/posts/danlor...

GitHub Actions Security Flaws: Immutable Tags, Token Access, and More | Dan Lorenc posted on the topic | LinkedIn

Yesterday I blamed the Trivy breach on GitHub. The design of Actions is plain irresponsible today and ignores a decade of supply chain security work from other ecosystems. Here's what they would have...

linkedin.com