Earlier this week I gave a talk to our internal hacking teams about the difference between good hackers and great ones that have been able to impact our field. I included three extended quotes - two from writers and one from Hamming that influenced my approach early on. Maybe some of it is useful for you.
Can sit and moan or make the best of it. I’ll be watching the match at the hotel. Not keen to watch in on my own. So lets give it a go and hope I get the “tech” sorted = TV on, My Laptop camera on, and you’re welcome to join me. Sign up here (adults 18+): watchwithrassie.com. This is personal, not official.
I have to admit that @dylfreed and @dnvolz at the @nytimes are amazing. Java4Shell is already amazing, but the article is going to make history for several other reasons:
1. For the first time, NY Times readers will be introduced to the phrase "heap grooming." Now that it is used in one of the world's most respected newspapers, perhaps @OED can add it to the dictionary.
2. It contains the full repro for the bug! The AI is going to use NYT as the harness now.
3. It contains "AAAAAAAAAAAAAAAAAA..."
I don't think it can get any better than this.
edition.cnn.com/2026/11/17/u…
This was a super generic public cyber piece until her advice! In years past it would have been “patch, update, strong passwords”. Now it’s “emergency cash, memorise phone numbers, emergency meeting spot”. WTF
The current state of AI is such that you could give five people in offensive security, all with varying skill levels, access to Codex or Claude and the exact same codebase or network to target, and get five very different results.
I spent weeks researching AI Pentesting Agents: can they be hacked? Yes. I got RCE on most Pentesting Agents ARTEX, strix, PentestGPT, etc. My research, stats, recommendations + 2 new concepts for Blue Teamers: TokenExhaust & HoneyAI are here:
plaintext.do/en/blog/hacking…
I must say even how stressful Q4 can be sometimes, there is nothing that beats being in the zone and doing awesome hacks into the night while blasting my favorite music on the headset.
Self-hosted Claude Managed Agents is now in private beta.
You can host the agent harness, sandboxes, and session history, in your own cloud or on-premises, and choose where requests to Claude are sent.
1) The vast majority of CVEs (vulnerabilities) are never observed to be exploited in the wild
2) A smaller percentage of CVEs are being exploited in the wild today than a year ago, probably because attackers haven't absorbed the massive new CVE supply
3) It's easiest to measure exploitation of CVEs that pertain to Internet-facing applications, since honeypot companies collect data on mass exploitation attempts and can see this as soon as these campaigns start (vs data collected after full on incidents)
4) Judging from this 'CVEs in internet-facing attack surfaces' view, it appears that exploitation of new CVEs is probably beginning to increase in the wild but the data are somewhat uncertain and this isn't happening in proportion to the new CVE supply
5) Hyperscalars like Amazon, Apple, Google are all good examples of vendors where there are new CVEs but we don't seem to be seeing mass exploitation; I think this is because they do a really good job (relative to the rest of software companies) of auto-security-updating their code and also spend, collectively > $1 billion on application security teams, tokens, and infra; so I think they're an anchor for how to do appsec in the AI security era
6) The long tail of the enterprise software security market is way, way weaker; fewer resources to do appsec (imagine zero or a few appsec engineers versus hundreds at the hyperscalars), software that doesn't auto-install security updates, companies owned by PE firms that've gutted security teams; this is where mass exploitation is happening
7) The core structural problem will be to find ways to use AI automation, and policy carrots and sticks, to bring 'long tail' enterprise software vendors up to something like the security hygiene of the hyperscalars
I've been pouring over these data for the past few weeks and still have substantial uncertainty, but I give a lot more depth here -- joshuasaxe181906.substack.co… -- plus a github repo showing how I attempted to kludge together a ton of partial, noisy data sources to estimate what's going on in vulnpocalypse-land.