Building Security + Enterprise Products @CoreWeave. Built Security Products @HashiCorp, Crypto and Defense Products @NetApp. 🖤: Kendo, trance music, Warhammer
NYC Metro Area
Joined August 2008
- Tweets17.6K
- Following669
- Followers2.1K
- Likes22.4K
Pinned Tweet
The traffic sucks. Everything's expensive. But living in a place where convos abt physics, CS, and math are the daily norm rules.
CYBERCOM and the like can engage KPA cyber adversaries because their kinetic deterrence consists of the most powerful military in history.
Cyberwarfare is still warfare. Showing up to a gunfight without guns? It’s a bold strategy, Cotton.
"There's interest in bringing back 'letters of marque and reprisal' to counteract the cybercrime that North Korea is committing."
North Korean cyber actors are military and intelligence personnel. If you're gonna go getting yourself into a direct confrontation with a hostile state's military apparatus, I'd say: gl, hf, and make sure you have a decent life insurance policy
Andrew "Andy" Manoske retweeted
For the first time in 10 years, a new major version of Postgres is delayed.... because AI is helping them find and fix so many bugs.
rhaas.blogspot.com/2026/10/p…
We've seen record sized breaches this past week and there has been a sharp rise in security incidents as AI is weaponized to find and exploit vulnerabilities in code.
What's less visible is the much steeper rise in how AI is being used to write more secure software and defend at every layer.
Builders have always had a fundamental advantage against attackers because they have full access to source code and control over system design.
We haven't always had the resources to fully take advantage of this asymmetry, but now with AI we do.
Everything we build can now be designed and pentested by the best security minds (models) in seconds -- for cents.
Not all teams have had the time to catch up yet, so we're not out of the woods as an industry.
But as the best teams fully modernize how they ship, AI is also enabling an era of the most trusted software ever. You don't see this in dramatized headlines yet, but it's a very good time to build.
RCE and root any Linux system with just a malformed packet.
AI is finding absolute demons hidden in the foundations of critical infrastructure.
What a disaster atlasum.net/blog/cve-2026-98…
Andrew "Andy" Manoske retweeted
I feel I need to explain this (I forgot people take me seriously): We can probably improve brute-force search at many places via reductions to tensor networks + OAI algorithm. Don't write a paper about each of them, just post them.
We’re upending foundational premises in math at breakneck speed thanks to frontier AI.
Some of those premises underly the security of common identity and encryption.
Not going to be surprised if we have to rebuild fundamentals of security + cryptography over the next decade.
Elliptic curves are not generic. Long time coming. Congrats @crypto_theory and collaborators github.com/bochum-crypto/sub…
What I said a week ago: It wouldn't surprise me in the least bit if there were a better factoring or discrete logarithm alg. For factoring: what is the chance that exp(n^1/3) is the right bound? who are we kidding? For discrete log over finite fields: better algorithms over Z_p have followed a decade or so behind those for small-char fields (Coppersmith'84 vs NFS a decade-ish later). Joux gave a quasi-poly alg over small-char fields in the 2010s. Porting to finite fields long overdue. For discrete log over elliptic curves: assuming that elliptic curve groups were "generic" is particularly preposterous given the rich structure they have. All of these ripe for disruption. Unstructured LPN would take me by real surprise, also unstructured LWE to a slightly lesser extent.
Short term this could be nigh apocalyptic if exploitations of proving P=NP or other earthshaking advances in math were utilizable in cryptanalysis.
But encryption will survive, and the beginnings of that are starting now:
nitter.cf/henryquantum/status/21…
Earlier this year, we experimentally demonstrated a digital signature scheme on Quantinuum's trapped ion quantum computer. This is a digital signature scheme that theoretically could be secure even in a world where P = NP. It necessarily requires a quantum computer to operate.
arxiv.org/pdf/2602.04859
The security of the scheme rests on a microcrypt assumption that it is computationally hard to reverse-engineer a random quantum circuit, based on its outputs.
Everyone who studied CS is leaning forward in their seats after reading the ETH news.
Cryptologists are doing the same, but far more anxiously.
What crazy times we live in.
Andrew "Andy" Manoske retweeted
So this is a silly story, but for a year or two before in 2013/2014 I went around telling everyone I could find about this little cryptography library called “OpenSSL” and how it was used everywhere, but had almost no developers or security budget.
Andrew "Andy" Manoske retweeted
Today we're taking the CoreWeave Global AI Network to the Middle East and Asia Pacific, on a backbone that runs at Petabit scale.
That's enough to stream a 4K movie to 66 million people at once. We'd rather move your datasets and checkpoints 🍿😉
All of it moves encrypted with MACsec, in hardware at line rate, across the backbone and inside our data centers, so security doesn't cost you throughput.
New points of presence in the UAE, India, Indonesia, Singapore, Australia, South Korea, and Japan are coming online beginning in Q4, bringing the network to four continents. crwv.co/uua8r
Andrew "Andy" Manoske retweeted
An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4
Andrew "Andy" Manoske retweeted
SANDBOX yourself, prevent this with a simple:
NetworkOptions(deny_egress=True, deny_ingress=True)
crwv.co/sandboxes
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
Tracking with everything here.
Employing software factories - and likely dedicated defense factories for blue team tooling that collaborates with the former - is going to become essential to secure enterprise computing.
Just spent the last 7 days on the front lines of the vulnpocalypse fending off an active cyber attack on one of our products.
Some thoughts:
- Software factories are going to be required equipment for everyone. There’s just no way to keep up with the pace of highly automated adversaries without one.
- We were lucky because we had spent the last few months getting our software factory running well. It is nice to have for product development but proved to be absolutely essential this week.
- At the minimum you need to have a coding server in the cloud pre-provisioned with read only access to all your logs and all the services you depend on for observability. With that in place you are a prompt away from reasonably good situational awareness, ongoing monitoring with loops, and ability for coding agents to validate fixes against your actual, observed attack patterns.
- Shared coding server also unlocks multiplayer incident response where everyone can see every session and jump in as needed. Sharing session links that can be forked is much higher bandwidth than re-explaining things.
- You don’t need to be Ramp/Stripe/Uber scale to have a working software factory in the cloud. We are a team of 3. Ours is built with an @opencode server (thanks @thdxr!), a forked version of compound engineering optimized for automated workflows (thanks @kieranklaassen!) and a multiplayer agent in slack that handles crons (we have our own harness but pick your favorite here).
- To get our software factory up and running well, we had to stop all new product development for ~6 weeks and just focus on getting the validation systems working that enabled us to reliably shipping PR without human code review. This felt very painful in the moment but was totally worth the eventual speedup in velocity.
- KYC for @AnthropicAI & @OpenAI cyber programs is now easy and quick (at least for defensive access), but a day felt like a lifetime this week. Submit your app now before you really need it.
- Getting Mythos access unlocked a deeper level of analysis that helped us shift from reactive mitigation to proactive patching. Wish we had done this a few weeks earlier. Get started now!
Andrew "Andy" Manoske retweeted
nice try
we spun up 10 THOUSAND sandboxes, @opencode running in each one pointing towards our serverless inference endpoint to get ASCII art
I'm feeling like an artist today
CoreWeave Scale
nice try
we spun up 700 boat.dev VMs 4c x 8gb, started codex + minecraft, made it all play in multiplayer together
they all booted from a snapshot in the first 10s, which cost $0.07, same price per VM, core & ram
but much more impressive capabilities and scale
The Eye of Sauron has fallen on Ruby. Multiple 9.0 (and even 10) CVSS vulnerabilities discovered today and luckily remediated.
Props to the community for patching so fast. Every major OSS project is probably on a hair trigger to respond right now.
ruby-lang.org/en/news/2026/1…
Well one of those crypto libraries now has problems.
The most popular DNS server in the world has a readily-exploitable RCE.
Not great.
kb.isc.org/docs/cve-2026-943…
4 mos ago I wrote a blog saying events like the last 48 hours in cyber were going to happen:
- We need to detect, respond, and degrade malicious agentic C2 infra
- We need to review common crypto libs ASAP. PHP2Shell on OpenSSH would be apocalyptic.
a2d2.medium.com/the-agentic-…
4 mos ago I wrote a blog saying events like the last 48 hours in cyber were going to happen:
- We need to detect, respond, and degrade malicious agentic C2 infra
- We need to review common crypto libs ASAP. PHP2Shell on OpenSSH would be apocalyptic.
a2d2.medium.com/the-agentic-…
On degrade: NSA's instructions are unprecedented. Not surprised if CYBERCOM is going Search and Destroy for malicious agentic C2.
It will likely not be enough, and we need to ask uncomfortable questions about how the private sector gets involved.
thehill.com/policy/technolog…
Example: Microsoft discovering of STORM-0940, a major Threat Actor leveraging agent swarms for the ConnectWise attacks.
Getting IoCs for their C2 is critical. But human IR is too slow for this adversary. When do we get proactive? What does that mean?
microsoft.com/en-us/security…
