frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Project Glasswing: what Mythos showed us

https://blog.cloudflare.com/cyber-frontier-models/
77•Fysi•2h ago

Comments

dataflow•39m ago
That's great and all but how severe were the most severe vulnerabilities found? I imagine they don't want to talk about it, but that's really the most interesting and important bit.
cute_boi•32m ago
Most of their new products are AI tools that nobody uses, so I guess they’ll keep posting slop. And recently, they’ve fired so many people that they probably don’t have good writers anymore.
aabhay•20m ago
As much as I’d like to share in the skepticism, the very beginning of the article states it very plainly — this is a step function.

Lots of people feel that Mythos is a psyops campaign, but I don’t really understand the skepticism. Most of it seems to stem from the general distrust of things that aren’t publicly available.

A few Anthropic employees have described Mythos as a general purpose model improvement, but that claim has yet to be widely backed up so that’s the only place I’m remaining skeptical.

For the domain of security research, I’m willing to buy the narrative.

ryandamm•14m ago
In his interview on the Hard Fork podcast, Palo Alto Networks’ CEO described the capability change from Opus to Mythos being more about availability; evidently it runs in a very compute-intensive, always-on mode. Unclear if the base model is significantly different, but Arora ascribed the difference mostly to that change.
xnorswap•34m ago
The real question is whether it was Mythos or Opus that wrote this post.

> "Why it matters"

It doesn't, it's a corporate blog, they were rarely written in one-author's voice anyway, but it's interesting to see that even large organisations are outsourcing their blogs to LLMs.

this_user•26m ago
This looks more like it was edited by AI rather than fully written by it. Or they are using a really good humaniser for the second pass.
estearum•23m ago
It's fascinating seeing people think that if you're snarky enough about something, the substance of that thing actually ceases to be substantive.

It's like staring down the barrel of a gun and taking the time to make quips about the type of paper the gun advertisement was printed on.

divan•22m ago
Cloudflare blogs have been excellent for many years, long before transformers arrived.
wnevets•29m ago
I can't wait to be told that Cloudflare is now part of "The Mythos FUD" campaign.
whizzter•9m ago
2 things can be true at the same time.

I think the curl folks finding it underwhelming is more of a testament to their code being subjected to a lot of tests/attacks/auditing over the past years compared to many other codebases. It's not going to find magically insurmounable exploits on it's own and "pwn teh w0rld".

At the same time, there is so much shitty non-memory safe code out there (C/C++ mainly) or logically weak code (much of it vibe-coded or otherwise by inexperienced devs) that will be easy pickings for anyone pointing Mythos at those codebases/services and eventually lead to chaos since the cost of an customized exploit has gone from days to months of expensive researcher time to some token spending.

Now if they noticed that they could find exploit chains easily in a lot of popular software, some embargo and hardening to give popular OSS packages time to not be exploitable by default does help people (and the NSA that probably has a preview).

sf_tristanb•26m ago
great, but why don't you share real data on how many security vuln it found ? how many were reals, how many weren't ?
hydra-f•24m ago
Beside the poorly written post, the vulnerability discovery workflow might actually give good results
unethical_ban•23m ago
Interesting for teams looking to implement ai into their deployment process.

I don't think guardrails are useful long term. Assuming we don't see the end of open near-frontier models, it is folly to try to keep models from doing exploit generation. The solution needs to be all software projects writing code under the assumption that hackers will be running LLMs against their code in search of exploits and write secure code accordingly.

MattSayar•17m ago
> The loudest reaction to Mythos Preview from other security leaders has been about speed - scan faster, patch faster, compress the response cycle. More than one team we have spoken with is now operating under a two-hour SLA from CVE release to patch in production [...] If regression testing takes a day, you cannot get to a two-hour SLA without skipping it, and the bugs you ship when you skip regression testing tend to be worse than the bugs you were trying to patch.

Over time, I wonder if these models will be able to generate more secure code by default by doing this kind of exploitability testing before ever merging their code.

wutwutwat•10m ago
Technically speaking CloudFlare is at its core, a security vulnerability itself. World's largest MITM
reducesuffering•8m ago
There will be no mea culpa from folks insinuating Mythos is a marketing stunt. Nor will there be every time AI capabilities repeatedly blast through the naive expectations.

We stopped AI bot spam in our GitHub repo using Git's –author flag

https://archestra.ai/blog/only-responsible-ai
94•ildari•57m ago•31 comments

Show HN: Files.md – Open-source alternative to Obsidian

https://github.com/zakirullin/files.md
243•zakirullin•2h ago•144 comments

1024000^2 Blocks, 2B2T Minecraft Server World Download Project, and Discoveries

https://github.com/2b2tplace/1m_release
61•exploraz•2h ago•31 comments

The Quiet Renovation at Bitwarden

https://blog.ppb1701.com/the-quiet-renovation-at-bitwarden
113•DaSHacka•1d ago•56 comments

Project Glasswing: what Mythos showed us

https://blog.cloudflare.com/cyber-frontier-models/
80•Fysi•2h ago•16 comments

A Rust-Python thing I am working on. Apache 2 licence

https://github.com/KevinKenya/nairobi-connector-open-source
5•kevinkenya•15m ago•1 comments

Voice AI Systems Are Vulnerable to Hidden Audio Attacks

https://spectrum.ieee.org/voice-ai-audio-attacks
30•SVI•4h ago•6 comments

The Aperiodic Table

https://blog.jgc.org/2026/05/the-aperiodic-table.html
48•jgrahamc•2d ago•12 comments

'We mould trees to grow into the shape of chairs'

https://www.bbc.co.uk/news/articles/cvg0yy3gp71o
123•bauc•3h ago•34 comments

Linux security mailing list 'almost unmanageable'

https://www.theregister.com/security/2026/05/18/linus-torvalds-says-ai-powered-bug-hunters-have-m...
134•jonbaer•4h ago•60 comments

It is time to give up the dualism introduced by the debate on consciousness

https://www.noemamag.com/there-is-no-hard-problem-of-consciousness/
212•ahalbert4•13h ago•531 comments

Porting my 3D points renderer on a ZX Spectrum 48K

https://github.com/ttsiodras/3D-on-a-ZX-Spectrum-48K/
55•ttsiodras•1d ago•8 comments

When Kierkegaard Got Cancelled

https://www.plough.com/en/topics/faith/discipleship/when-kierkegaard-got-cancelled
23•bookofjoe•4h ago•5 comments

GenCAD

https://gencad.github.io/
402•dagenix•18h ago•107 comments

Enough with the AI FOMO, go slow-mo, says Domo CDO

https://www.theregister.com/ai-ml/2026/05/17/enough-with-the-ai-fomo-go-slow-mo-says-domo-cdo/524...
107•Bender•3h ago•51 comments

Show HN: Auto-identity-remove – Automated data broker opt-out runner for macOS

https://github.com/stephenlthorn/auto-identity-remove
298•stephenlthorn•4h ago•117 comments

The foundations of a provably secure operating system (PSOS) (1979) [pdf]

http://www.csl.sri.com/users/neumann/psos.pdf
86•rurban•6h ago•49 comments

Crystals found inside wreckage from the first nuclear bomb test

https://www.scientificamerican.com/article/strange-crystals-found-inside-wreckage-from-the-first-...
142•jumploops•2d ago•64 comments

What Is Date:Italy?

http://aesthetikx.info/blog/date_italy.html
4•jollyjerry•1d ago•1 comments

Math Jokes in Alice in Wonderland

https://storica.club/blog/alice-is-math-jokes/
86•yekbun123•4d ago•40 comments

Don't answer the first question

https://lalitm.com/post/dont-answer-the-first-question/
38•lalitmaganti•7h ago•27 comments

Learn Harness Engineering

https://walkinglabs.github.io/learn-harness-engineering/en/
11•redbell•4h ago•0 comments

Ask an Astronaut: 333 hours of Q&A footage with astronauts

https://askanastronaut.issinrealtime.org/
183•gaws•2d ago•22 comments

Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep

https://github.com/MinishLab/semble
400•Bibabomas•1d ago•133 comments

AI eats the world (Spring 26) [pdf]

https://static1.squarespace.com/static/50363cf324ac8e905e7df861/t/6a0af5d0484fbf5fe9a7743e/177910...
134•topherjaynes•3h ago•73 comments

Jank now has its own custom IR

https://jank-lang.org/blog/2026-05-08-optimization/
195•DASD•2d ago•37 comments

What “Amazon Supply Chain Services” Tells Us About What Amazon Is

https://gadallon.substack.com/p/the-third-time-amazon-did-this-what
15•JumpCrisscross•1h ago•6 comments

Build a Radio Wave Detector with Balls of Aluminum Foil

https://www.wired.com/story/build-a-radio-wave-detector-with-balls-of-aluminum-foil/
51•Brajeshwar•2d ago•16 comments

Researchers Wanted Preschool Teachers to Wear Cameras to Train AI

https://www.404media.co/researchers-wanted-preschool-teachers-to-wear-cameras-to-train-ai/
22•cdrnsf•2h ago•1 comments

NASA still maintains some of the Voyager spacecraft code from the 70s era

https://spacedaily.com/nasa-still-maintains-some-of-the-voyager-spacecraft-code-in-a-1970s-era-pr...
64•redbell•5h ago•63 comments